Datasets:

ai-aerospace

/

ams_data_full_2000-2020

like 0

189 A post -life teardown of the gearbox revealed the bearing races showed signs of use, however nothing atypical was observed. On both inner and outer races of the roller bearings, the contact zone was evident by a visibly darker band visible in Figure 11. The grease within the PEEK cages around the rollers was darker, but still wet with adequate lubricant for proper operation. The PEEK cages appeared in good condition as well, maintaining their integrity. STIG Testing Results The assembled qualificatio n model corer has completed a 1x life test program of actuator/gear train rotations and gear shifts. At the time of writing, the unit has also performed approximately 20% of a life’s worth of coring and drilling operations. This testing will continue. The flight model corer has completed dyno testing, completed thermal -vac testing, and has been integrated to the flight rover. There it has performed basic operations, such as shifting gears. Figure 12 shows performance data for the f light unit taken across the protoflight temperature range during testing. >100% margins were proven for dril ling and core breakoff operating points. Drilling operations were performed while connected to a dynamometer . Core break operating points wer e measured quasi -statica lly against a load cell, since that operation can be performed arbitrarily slow in flight. Challenges and Lessons Learned Several challenges were encountered during the design of STIG. Shift Energy Compensation Spring preload allows for shifting into gear without pre -alignment of selector and actuator output splines . However, energy on the order of 0.3 J is stored and released once the splines align rotationally. This energy manifests in the form of axial velocity of t he selector of about 2.5 m/s . Accommodations must be made for this kinetic energy; allowing the selector to dissipate energy via impact with the actuator output interface (specifically , at the termination of the depth of the selector’s internal splines) was not acceptable. Damping was considered, for example with piston seals, but the large range of interference fits that would be encountered over the operating temperature range of the mechanism on Earth and on Mars and over the life of these seals meant th at constant friction coefficients could not be guarantee d. A compact absorber spring of high stiffness was designed and incorporated in order to release energy into other STIG parts with a load path to corer structure. The absorber spring, along with a few small parts contained within it, surround the selector and can be impacted without blocking the selector’s interfaces to the spindle actuator. This particular load path also included the transmission’s output bearings, whose truncation loads of 3 kN became the limiting factor for maximum impact forces. The design of the absorber spring was made difficult by the need to absorb a fixed amount of energy in a small packaging volume. As the spring geometry became more complex to reduce stresses, increases to tolerances were necessary to allow the part to remain machinable. Given that the spring needs to flex in order to absorb energy, stresses are quite sensitive to tolerances in part thickness. A great deal of iteration was used in the design of the absorber spring. Figure 13 shows the absorber spring final design, which stores energy through torsional and radial flexing. Qualification model testing has shown the bearings to be capable of sustaining the 3 kN max loads without impacting bearing performance.

Document

AMS_2020.pdf

190 Figure 12 . Torque and current data from the corer in the high -speed gear at drilling speed (top) and in the high torque gear measured statically against a load cell (bottom), for positive and negative directions of spindle rotation. Data was collected at and between - 70 °C and +50 °C, the protoflight temperature requireme nts of the corer body based on the Jezero Crater landing site.

Document

AMS_2020.pdf

191 Figure 13. The absorber spring has a height of about 25 mm and a diameter of about 19 mm . The bottom end is threaded for installation into the output shaft that surrounds the selector . The top end includes a feature where a two- piece split collar and retaining ring are placed . Contact with the selector is made at this collar during impact . Regions of highest stresses are found at the round holes that terminate each slot. When designing a transmission that utilizes spring preload to drive gear changes, it was found to be essential to study all off -nominal releases of energy and to plan for shift s occurring prior to planned state changes. For example, lead- in chamfers on the selector, designed to aid in engagement of splines during shifting, made it possible for the selector to rotate and backdrive the spindle actuator due to axial forces sustaine d during shifting. Were backdriving of the selector to occur before the STIG lead screw reached its end of travel, the selector could be accelerated and small par ts used to load the STIG spring w ould see impact loads against each other. These parts demanded a great deal of structural analysis and design iteration before this load case was deemed to be safe. When shifting in the high torque direction, the selector impacts corer structure. Although the compliance of that structure is less than that of the absorber spring, the direct path of the load to structure means that acceptable loads were substantially hi gher. Therefore, designing parts for impact here required less iteration than that of the absorber spring. Shunting loads to structure is preferable to a load path that includes small or fragile mechanism components whose designs are driven by other requir ements. Force Uncertainty in Coupled Lead Screws An additional lesson learned was the need to plan for large force uncertainty when using two lead screws geared together and driven by one actuator . Due to large differences in lead screw diameter (6 mm vs 31 mm, for the CBLO and STIG lead screw, respectively), the linear force output to input torque relations of the lead screws are quite different. Potential small changes in observed actuator current could be due to noise; overcoming a small amount of unpl anned drag in the large, low -output force ratio lead screw; or forces of several kilonewtons in the smaller, high- output force ratio lead screw. While testing a prototype

Document

AMS_2020.pdf

192 concept, a piece of ground support equipment that did not provide clearance for the f ull range of travel of the high output CBLO lead screw nearly led to the generation of forces capabl e of severely damaging the corer . This fault was prevented by an operator E -stop press. The late addition of a spring in the CBLO mechanism allowed for these high forces to be eliminated in a specific jamming fault case of concern. Conclusions When coupled to a dual -output actuator, STIG provides a compact approach to gear shifting. In the case of the corer, by leveraging the position of a different mechanism, STIG does not require the addition of an additional actuator specifically tasked with gear selection. STIG, through the use of spindle actuator stalls, provides feedback that gear shifts have completed and splines are in full contact. No additional sensors are required. STIG retains the ability to provide motion in either direction in eithe r gear, by avoiding the use of ratchets and impact drivers . When designing mechanisms similar to STIG that use a spring to allow gear shifts to be agnostic to actuator output spline positions, care must be taken to plan for energy dissipation. Coupled lead screws of different sizes are an additional challenge, due to their tendency to obscure large force generation. Testing of STIG, along with the core break lockout mechanism continues in the qualification model test program at JPL. Launch is pl anned for July of 2020. Acknowledgement Except where noted otherwise, this work was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Spac e Administration (80NM0018D004) . © 2020. All rights reversed. References 1. Myrick., T., US Patent No. 6,550,549, "Core Break off Mechanism” 2. LoSchiavo et. al . “Mars 2020 maxon Commercial Motor Development from Commercial -Off-the-Shelf to Flight -Qualified Motors, Gearboxes, and Detent Brakes: Overcoming Issues and Lessons Learned” . Proce edings of the 45th Aerospace Mechanisms Symposium. 2020. 3. Chrystal, Kyle . “Percussion Mechanism for the Mars 2020 Coring Drill” . Proce edings of the 45th Aerospace Mechanisms Symposium. 2020. 4. Lo, C.J., et al., “Use of Cumulative Degradation Factor Prediction and Life Test Result of the Thruster Gimbal Assembly Actuator for the Dawn Flight Project”, NASA CR -2009- 215681 5. Conley, P.L. and Bohner, J.J., “Experience with Synthetic Fluorinated Fluid Lubricants”, Proc. 24th Aerospace Mech. Symp., NASA CP- 3062, Kennedy Space Center, FL (1990)

Document

AMS_2020.pdf

193 Development of the Next Generation Battery Cell Isolation Switch Ruben Betancourt* and Michel Knight* Abstract EBAD’s NEA Battery Cell Isolation Switches (BCIS) have been used to isolate the electrical circuit of a lithium -ion cell within the battery due to safety or performance concerns . Previous testing has shown that there are limitations to the capability of the heritage design under extreme thermal environments and low actuation input current . For example, if a BCIS was functioned at low temperatures and low actuation current there was a possibility of the device not meeting the actuation requirements. This limitation led to the design of the next generation BCIS as described in this paper. The next generation NEA BCIS was designed to be able to operate under the most extreme environments and lowest current that our customers have requested . This was validated with a combination of analysis and testing. Introduction BCIS Design Description The BCIS is an electromechanical switch that serves two basic functions. The first is that it is used to interconnect lithium ion battery cells , which requires the BCIS to conduct a continuous current up to 400 amps. EBAD has three BCIS product lines: the NEA8020 series with a 135- amp current carrying capability, the NEA8030 series with a 250- amp current carrying capability, and the NEA8040 series with a 400-amp current carrying capability. The different models operate in a similar manner but are scaled depending on the current carrying capability. A cross section of a typical 8030 BCIS is shown in Figure 1. The components on the right side are the high current carrying components which conduct the power output by the batteries. Prior to activation of the BCIS, the electrical terminals T1 and T2 form a closed circuit and conduct the high current, while T3 remains electrically isolated. The second function of the BCIS is to isolate a battery cell from other batter y cells. When a battery shows signs of degradation, the BCIS may be utilized to divert the flow of electrical power between sets of terminals to isolate the degrading battery cell . On the left side of Figure 1 is the release mechanism portion of the BCIS, which provides the switching mechanism. The BCIS utilizes a set of internal spools which are restrained by wire, which is then attached to a fuse wire (T4 to T5) . The switching function is initiated by the application of a minimum activation current of 1.5 A for a duration of up to 230 ms, across the fuse wire circuit . The current causes the fuse wire to break and release the restraining wire that ho lds two spool halves together. This allows a preloaded spring assembly to push a plunger forward , which creates a closed circuit between T1 and T3, and electrically isolates T2. Related to this switching function, the BCIS must meet a set of requirements which include: • Commutation time less than ~10 msec (varies by customer) o Defined as the time required to sw itch the circuit from T1 -T2 to T1- T3 • Make- Before -Break time of <1 ms o Defined as the time all electrical terminals (T1, T2, T3) are in electrical contact * Ensign- Bickford – NEA Electronics Inc., Moorpark, CA Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

194 Figure 1: NEA BCIS 8030 Family Design Development Background At the 2019 European Space Mechanisms And Tribology Symposi um (ESMATS), a summary of lessons learned during Qualification testing of the heritage design was presented. The presentation discussed how the BCIS had failed to actuate properly , and the details can be found in the ESMATS archive (Titled “ Bypass Switches – A Case Study in Test Like You Fly Merits ”). To summarize the results, t he failure investigation found that there were multiple causes for this anomaly. 1. A low actuation current (less than 3 amps) leads to an undesirable condition that could affect the performance of the BCIS. 2. At temperatures less than ambient, the likelihood of an anomaly increases. The reason that these two causes led to a failure was because they led to a potential negative force margin on the split spool actuation device within the BCIS . Force margin is defined as the ratio between the separating force and the resistive force. In this case, it is specifically the ratio between the restraining wire separating force and the fuse wire res istive force, including all frictional forces . If the restraining wire separating force is greater than the resistive forces caused by the interaction between the restraining wire loop and fuse wire, then the restraining wire can unwind resulting in successful actuation. The heritage design used a restraining wire with a 0.014- in (0.36- mm) diameter , which corresponded to a separation force of approxi mately 0.44 lbf (2 N) . In most cases, this force was sufficient to separate the restraining wire from the fuse wire, leading to a successful actuation, but the causes 1 and 2 listed above, combined with production lot variation, led to a possible condition where the restraining wire would not separate from the fuse wire or the separation would be delayed. This paper discusses the design iterations and lessons learned during the design and qualificatio n of the improved BCIS design, specifically within the release mechanism portion of the BCIS . Design Development The development of the improved BCIS started with a set of basic requirements: 1. The design improvements should not affect the envelope of the existing desig n 2. The improved design should be able to withstand the most extreme environments and conditions that have been requested by our customers 3. The electrical performance of the BCIS should remain unchanged Spool Restraining Wire Fuse wire Plunger High Current Carrying Components T2 T3 T1 Release Mechanism T4 & T5

Document

AMS_2020.pdf

195 Knowing the limitations of the existing design, the app roach was to improve the force margin to such an extent that low current, environmental impacts and assembly variation would not affect the performance of the BCIS. There were two initial designs that were considered. The first was to reduce resistive forces by remov ing some of the insulating material around the fuse wire and restraining wire. This would eliminate frictional forces and would guarantee a positive force margin. However, a Design Failure Mode and Effects Analysis ( DFMEA) was held and multiple potential risks were identified. Removing material had the potential of creating pinch points, snagging possibilities and other potential risks that eventually led to the elimination of this option. A rigorous vetting process was abl e to detect these issues before significant resources were expended on this design. The second option was to increase the separation force. The easiest approach was to increase the restraining wire diameter, there by increasing the spring back force of the wire and yield positive margin. Initially, two different sizes were considered, a 0.016- in (0.41- mm) diameter wire and an 0.018 -in (0.46- mm) diameter wire. A nalysis and prototype testing were performed to determine the best choice. Proto type Testing The spring back force of each size restraining wire was calculated through a mathematical model, and both sizes were expected to provide significant margin over the heritage design. Since the envelope of the design was critical, an assessment of the clearance was one of the first tests performed. Multiple units were functioned, and the unwinding diameter of the restraining wire was measured. This w as important, as a large unwinding diameter could indicate that the restraining wire may not have sufficient clearance to unwind properly . In addition to post -test inspection, functional testing was performed with the use of a hi ghspeed camera. It was clear that the 0.016- in (0.41- mm) restraining wire unwound smoothly and fit well inside the BCIS, while the 0.018- in (0.46 -mm) restraining wire unwound to a considerably larger diameter and resulted in a condition where the restraining wire could potentially become jammed against the wall of the housing. Therefore, the 0.016- in (0.41- mm) restraining wire was chosen for the final design. The analytical model was validated by test and shown to be in good agreement with the test results, with the restraining wire loads found to be normally distributed about the mean. It was also shown to be a si gnificant improvement to the heritage design. Figure 2 is a plot of the restraining wire force of both the 0.014- in (0.36- mm) and 0.016- in (0.41- mm) restraining wire. The data shows that with the 0.014- in (0.36- mm) restraining wire there exists a very low probability of not providing enough force to separate from the fuse wire. For the 0.016- in (0.41 -mm) restraining wire, t he force margin was always positive when using a +4 sigma for the frictional force and a -4 sigma for the separation force. Figure 2: NEA Ultimate Load Test

Document

AMS_2020.pdf

196 While performing the prototype testing with the 0.016- in (0.41- mm) restraining wire, the BCIS was tested to very extreme conditions to draw out any other potential issues. This included using various actuation currents, test temperatures, and vibration levels. Based on this testing, the next limiting component in the design was the spools. The heritage design used a spool that was made of a glass -filled Torlon and reacting against it was a metallic plunger with an edge that only had a small radius. Due to the low loads on the plunger, this was not seen as an issue as there were never any signs of indentations or any other observable issues . However, when testing at extreme environments, the additional wear from higher level environment and temperatures , revealed a new failure mode . High-speed video revealed that the restraining wire could fully unwind, and yet the spool would temporarily not move ( Figure 3) . While this was only temporary, it would fail to meet the customer commut ation time requirement. Figure 3: Unwound Restraining Wire with No Movement From the hi gh-speed video it was evident that the slow commutation time was related to either the spool - to-insulator interface, or the spool -to-plunger interface. It was determined that the lowest risk approach would be to replace the plas tic spools with metallic spools. This would create a better coefficient of friction between the three components and would eliminate the concern of any spool indentations. Since the load on the components was low, Aluminum 7075 was chosen. A DFMEA was held to determine potential risks with a material change. The biggest concer n was that the metallic spools would be conductive, which could lead to a short between the terminal T3 and the actuation circuit, as seen in Figure 4. Figure 4: BCIS Potential Electrical Path In order to eliminate this concern, a thin wall of insulating material was added to the insulating driver . This added a physical barrier between the metallic plunger and the metallic spring that contact s T3. In addition, Restraining Wire Fully Unwound No Plunger Movement

Document

AMS_2020.pdf

197 a proprietary coating called Magnap late® HCR was applied to the spools. This provided three benefits. The first is that Magnaplate® HCR has high dielectic strength, meaning that it will isolate the metallic spools from any other metallic components . The secon d benefit is that Magnaplate HCR increases the hardness of the aluminum spools to above 50 Rockwell C , which is harder than the plunger. It is ideal for the spools to be harder than the plunger, as this prevents the plunger from digging into the spools and preventi ng or slowing release. A stress analysis was performed and showed that the stress remained well below the material capability for the spools and plunger (Figure 5) . Testing was also performed to validate the results. The testing showed that the spool and p lunger combination could sustain loads over 900 lbf (4000 N) , while the load during use was expected to be on the order of 44 lbf (2 N) . Figure 5: FEA of the Spool and Plunger The final benefit is that Magnaplate® HCR is self -lubricating , with a published coefficient of friction of 0.35 when used in combination with aluminum. Since the plunger is stainless steel, it is likely that the actual coefficient of friction between the plunger and Magnaplate coating is less. A cros s-section of t he final design is shown in Figure 6. The design meets t he primary design requ irement of fitti ng within the initial envelope and it has passed all initial development tests. Figure 6: Final BCIS Design Test Results As of the end of 2019, the unit has successfully completed prototype and development testing. The development testing included the following: • 40+ actuations o Actuation Current = 1.5 amps and greater

Document

AMS_2020.pdf

198 o Temperature = - 70°C to +130° C • Vibration Testing : o Sine Vibration: Up to 110G o Random Vibration: Up to 56 Grms • Shock Input o 2300G from 1300 Hz to 10000 Hz • Thermal Shock: o -62°C to +137°C o 10 cycles Qualification testing will be conducted in April 2020 and is expected to be completed by May 2020. The expected qualification sequence is in Table 1. Table 1: Qualification Test sequence Conclusions and Lessons Learned 1. EBAD has developed an updated NEA BCIS design that is insensitive to extreme environments and actuation conditions (within reason) 2. During the initial phase of a program, parallel paths should be chosen to reduce the impact of unforeseen risks 3. Perform mini design review s, DFMEAs, etc. early and often to prevent spending resources on risky design concepts 4. The force margin on even the smallest components should be determined

Document

AMS_2020.pdf

199 A Fast -Acting, Self -Energized, Low -Cost Valve for Air Cannons Lee E. Brown* Abstract A novel pneumatic valve has been developed to control pressurized air for an air cannon. It offers very rapid opening time, is self -energized when opened, self -locking when closed, and is of relatively low cost. The valve uses a variation on the classic four -bar linkage which enables the valve to be self -locking against pressure when closed, yet self -energized and opens very rapid ly when the valve mechanism is tripped. The valve has been tested and has been us ed in the field to test munitions fuzes . Background Air cannons have long been used for dynamic testing of electronics and mechanisms. These air cannons are used to simulate high- G environments such as artillery fire or high- speed impacts. They can also be used to propel electromechanical devices past objects of interest, such as testing munitions fuz es against specific types of targets. The use of air instead of chemical propellants greatly simplifies these kinds of tests and greatly reduces the cost of such testing. The key component in any air cannon is the means by which the high pressure air is controlled. A very rapid rise in pressure is needed to accelerate the projectile to the desired muzzle velocity within the shortest possible barrel length. The oldest means of obtaining a very rapid pressure rise is through the use of a burst disk. A burst disk is a membrane designed to fail at a specified air pressure and thus achieve a sudden release of air into the cannon’s chamber . Burst disks are capable of controlling very high pressures and are usually designed with scoring or grooves that give fairly repeatable failure pressures. The drawback to burst disks is that they must be replaced after each shot which is time consuming. High pressure burst disks are often precision machined elements which adds to the cost per shot. At lower pressures metal or plastic foils are employed; however , they tend to operate with less consistency and thus higher variations in projectile velocity. More recently, electromechanical pneumatic valves have been tried in lieu of burst disks. These valves avoid the per -shot expense of burst disks but the initial purchase price of valves capable of handling high pressures can be quite high. Electromechanical valves also offer substantially longer pressure rise times than do burst disks and longer barrels are necessary unl ess some sort of ret arding mechanism is used to hold the projectile in the chamber until peak pressure is achieved. Requirements In the spring of 2015 Electronics Development Corporation undertook the development of a 75-mm bore air cannon to be used for t he testing of munitions fuzes. The design goal for the 75- mm air cannon is to propel a 1- kg projecti le with a muzzle velocity of at least 75 m/s. It must use the shortest possible barrel length and be very easy to operate. Low cost, both for cost of acquisition and cost per shot is also a factor. * Electronics Development Corporation, Columbia, MD; lee.brown@elecdev.com Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Center, 2020

Document

AMS_2020.pdf

200 It was decided that this air cannon should use a portable air compressor or “shop air” in the range of 400 to 800 kPa. Early experimentat ion with burst disks showed that metal foil disks yielded erratic results at these pressures and they tend to shed foil debris which cause s serious problems for some types of electronic fuzes. Conventional electromechanical pneumatic valves were ruled out due to slow opening times. Given the drawbacks noted above, Electronics Development Corporation elected to develop a novel type of pneumatic valve that offers fast pressure rise times comparable to burst disks yet be simple and inexpensive. This valve is self-locking against pressure when closed yet be self -energized when opening. Design The schematic design of th e air canon is shown in Figure 1. Pressure is supplied by a 75 L pressure vessel with a 50 mm diameter opening. The valve inlet a nd outlet apertures are also 50- mm diameter. Downstream of the valve the aperture increases to 75- mm diameter for the breech and the barrel. Figure 1: Schematic Arrangement of air cannon The valve’s components are arranged as shown in Figure 2. The valve consists of an inlet and an outlet aperture, a flapper element, the flapper linkage, and an impact absorber. A pneumatic actuation cylinder and a trip lever are mounted on the rear of the valve and may be seen in Figure 3.

Document

AMS_2020.pdf

201 Figure 2: Physical arrangement of valve In order for the valve to achieve the self -locking and self -energizing requirements a design variation on the classic four -bar linkage was chosen as shown in Figures 4 and 5. This linkage (a-b-c-d) goes over center (angle b- c less than 180o) to lock against the pressure force but only requires a small external torque to trip Figure 1: Resetting the valve the mechanism out of the over center lock whereupon the pressure rapidly throws the valve open. An impact absorber is provided to stop the moving linkage when opened. If the over center angle is kept small very little torque is required either to trip the valve or to reset it. The torque to actuate the valve is supplied by a small pneumatic cylinder attached to the back of the valve. This version of the valve is simply reset by hand as sh own in Figure 3. Impact Absorber Linkage Flapper Inlet Aperture Outlet Aperture Pressure Transducer

Document

AMS_2020.pdf

202 Figure 4: Valve in Closed position Figure 5: Valve in open position

Document

AMS_2020.pdf

203 Testing A sample pressure curve is shown in Figure 6. Of note is the rapid pressure rise time of ~10 ms. While this is not quite as fast as a burst disk it is much faster than the majority of electromechanical pneumatic valves of similar aperture. The 75- L pressure vessel completely empties in ~200 ms. Figure 6: Sample pressure curve Live fire tests with 1- kg dummy test projectiles showed that the desired muzzle velocity of 75 m/s is easily achieved at ~250 kPa pressures using a 3- m barrel. 100 m/s is possible using the full 400 kPa max operating pressure. Field Use The first field use of the finished 75 mm air cannon was testing prototype munitions fuzes by mounting them into test projectiles and shooting them past targets of interest . Figure 7 shows one of the test setups (note the barrel in the foreground.) This testing required hundreds of shots to be fired in as rapid a succession as possible. The valve held up extremely well with no misfires or other mechanical deficiencies. No adjustment s to the valve were required throughout the testing.

Document

AMS_2020.pdf

204 Figure 7: Use in the field Conclusions The valve for the 75- mm air cannon has met or exceeded all requirements. The over center four -bar linkage is surprisingly robust , needing no adjustment over the course of hundreds of shots. The valve can be manufactured at a relatively low cost as the parts are simple and few tight tolerances are required. It bears noting that this valve design is not limited to use on air cannons. The design is easily scalable and could potentially be used in any application which requires t he rapid venting of a pressurized fluid.

Document

AMS_2020.pdf

205 Development of a L ow-Shock Separation Nut Out of the P yrotechnic Class Sebastien P erez*, Frederic Miralles * and François Degryse* Introduction About Pyroalliance Pyroalliance is the European leader and a world- class player in pyrotechnics and mechanisms, delivering innovative and cost -effective solutions tailored to our customers’ critical requirements. With more than a half-century of experience as a manufacturer of state -of-the-art pyrotechnic and mechanical equipment, Pyroalliance addresses the requirements of cutting- edge industries, including Aerospace, Defense and Energy. We leverage our proven expertise and innovative mindset to develop products that combine the performance and reliability needed to perform critical functions for our cust omers’ systems. Facts and figures: • Counting over 240 collaborators, Pyroalliance is operating from two locations. The headquarter s is at Les Mureaux (Region of Paris) and the second site is located in Toulon (Southe ast France) • Pyroalliance is 90% owned by ArianeGroup, the remaining 10% be longing to the company OEA Inc. • The company has reached 40 million Euros turnover in 2018, representing a 15% growth compared to 2017 • Exporting in more than 15 countries, Pyroalliance is pursuing its international growth Low-Shock Separation nuts By adapting the design of its Pyrotechnics Nuts, Pyroalliance overcomes the preconceived notions and changes the game by claiming its products are Ultra Low Shock. Pyroalliance has been desi gning and producing Pyrotechnic Nuts as well as Hold and Release Mechanisms for decades. Those are dedicated to maintain satell ites on dispensers as well as antennas and booms on satellites. To date, Pyroalliance has delivered more than 6000 of those with a track record of 100% operational success. Indeed pyrotechnic nuts provide major operational benefits thanks to their very high energy density, very high standard of proven reliability, speed of execution (milliseconds scale) and ideal synchronicity when several nuts are to be activated simultaneously (typical situation when satellites are h eld on a satellite dispenser). Recent market surveys show that: • Separation Nuts require a growing capability to reduce levels of shocks on the payloads, what is considered as an attractive feature for sensitive payloads, combined with an ability to ensure synchronicity; • Despite of their performance, a growing number of actors of the New Space are reluctant to use equipment classified as pyro for operational reasons. Understanding those challenges , Pyroalliance presents hereafter its recent achievements in those areas . * Pyroalliance, Les Mureaux, France; sebastien.perez@pyroalliance.com, frederic.miralles@pyroalliance.com, francois.degryse@pyroalliance.com Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

206 Separation Nut Background Pyroalliance designs , manufactures and delivers separation nuts or complete Hold and Release Mechanisms meet ing the requirements of various mission profiles . However, each design is based on the same principles which have cumulated decades of flight heritage . Figure 1. Pyroalliance separation nuts – Principle Functioning principle : The pyro- initiators make the pressure increase inside the sealed chamber, until the release piston can translate. When shifted, the release piston, due to its inner geometry, allows the radial expansion of the thread segments and, consequently, the release of the s crew. At the same time, the ejection pin moves and ensures that the screw cannot remain or move back to its initial location. Pyroalliance separation nuts are fully reset table on both Pyroalliance sites or customer facilities using cold gas (non pyro gas ). The Spring allows reset ting the parts in the initial position when actuated by cold gas. Before delivery , 100% of our separati on nuts are controlled using cold gas actuation in a test set reflecting the extreme operational conditions of the mission (temperature and pre -load at release). The heritage design has the following main advantages: • Outstanding reliability figures (> 1 – 5x10-5) despite extreme operational conditions ; the simple design made of only a few parts and the motorization margins driven by the pyrotechnic s are the key features enabling such level of reliability • Extremely low actuation time / actuation time standard deviation: the actuation time of Pyroalliance separation nuts remains below 2 ms cumulating both contributions of pyrotechnic and mechanical actuations . Then, the actuation time standard deviation is well below 1 ms which is particularly suited for multi -point release. On top of those advantages, Pyroalliance recently started to implement some adaptations to this heritage design in order to meet new emerging market requirements . They aim at : • Reducing the level of shock induced on the holding structures during the releas e • Performing the necessary steps to make the device Out of Class (non pyro). Pyroalliance is curently working o n implementing those adaptations while preserving all assets of the heritage design. Release pistonCombustion chamber Thread segmentEjector SeparatorSpringHousing BaseInitiator

Document

AMS_2020.pdf

207 Pyroalliance achievements First analysis When analyzing the separation nut heritage design, Pyroalliance made the assessment that there were three main contributors to the induced shock : • Pyrotechnic shock due to the initiator actuation • Strain energy stored in the screw during the preload and freed during the release • Kinetic energy stored in the release piston during the release and further distributed during following shocks of the piston with other mechanical parts . Table 1 summari zes the different contri butor s of the shock and quantif ies their respective impact. The considered design is the heritage M10 separation nut (screw diameter: 10 mm / 0.39 inch) . Table 1. Shock contributors assessement – M10 separation nut heritage design Shock source Nature Shock contribution* Initiator actuation Pyrotechnic 10% (tested with NASA Standard Initiator / Space Standard Initiator ) Screw release of the preload Mechanical 5% (preload: 20 kN ; screw length : 100 mm ; screw material: stainless steel) Impact of the release piston Mechanical 85% *Note: Several tests / ana lysis have been performed in order to quantify the impact of each contributor. Pyroalliance found out that the shock was mainly caused by the impact of the release piston at the end of his stroke once the threaded segment s are expa nded. As an ex ample, the sketch on Figure 2 presents the c haracteristic of the piston collision using the M10 separation nut . The velocity of the piston at the time of the impact has been captured using a high- speed camera. Figure 2. M10 separation nut – Collision characteristics Impact characteristics •Release piston mass: 50g •Release piston velocity: 30 m/s •Kinetic energy: 22.5 J

Document

AMS_2020.pdf

208 Addition of a damping system Based on those findings, Pyroalliance designed a specific damping system made of a dedicated damping material and located after the release at the end of the release piston stroke. This system allows performing a progressive braking of the release piston without degrading the reliability of the global syste m. Furthermore, this damping system is fully compliant with respect to LEO space requirements: • Outgassing compliant with NASA / ESA Standards • Tested on the temperature range [ -60°C;+80°C] without any performance degradation • Compatible with a 10 years storage duration without any performance degradation (tested through accelerate d aging ). Pyroalliance has performed several tests with the addition of the damping system. The shock configuration is composed of an aluminum plate of 1m×1m× 5mm receiving the separation nut test assembly. Four accelerometers are distributed on a 100- mm-diameter circle around the separation nut. Figure 3. Shock test configuration – M10 separation nut

Document

AMS_2020.pdf

209 The results are presented on the Figure 4. Figure 4. Shock Response Spectrum - Addition of the damping system These results have been measured o n Pyroalliance own shock test lab configuration. A comparison of the shock with / without damping system has been performed in the exact same configuration considering several units. The induced shock has been reduced by a factor 4 on the frequency band [1000 Hz; 10000Hz]. Furthermore, the damping system brings similarly better results on the entire frequency range. Tests have also been performed on a spacecraft mockup at customer level . The accelerometers were located on the spacecraft side close to the cup / cone interface . The results measured by the customer were the following: - 4000 g SRS @2000Hz with the heritage design (without damping); - 1000 g SRS @2000Hz with the addition of the da mping system. The shock reduction by a factor 4 measured on the shock test lab configuration has therefore also been verified with the same ratio at customer level. Pyroalliance is pursuing this effort to improve the damping system and reduce the shock induced by the separation nuts with a target of 500 g SRS @2000Hz on a spacecraft configuration . Out of classification European procedure In the context of aerospace industry, Pyroalliance is turning toward techn ologies which contain pyrotechnics and its condensed power while being as safe as other non pyrotechnic devices. In order to be classified as non- dangerous good regarding the law (c.f French transport regulation law: Recommandations relatives au transport des marchandises dangereuses ST/SG/AC.10/1/Rev.20) , Pyroalliance will conduct several tests to demonstrate that an unpackaged separatio n nut equipped with pyro- initiators satisf ies the 6 criteria in Table 2. 1·1011·1021·1031·104g 21·1031·1041·10 Hz SRS MAXIMAX (Q=10) - Shock comparison - M10 separation nut separation nut without damping system separation nut with damping system

Document

AMS_2020.pdf

210 Table 2. European criteria – non dangerous goods transport regulation Criterion Content Status Sensor type Property when functioning 1 Surface temperature When actuated, external surface s of the product shall not reach a temperature higher than 65°C (149°F). A transient 200°C (392°F) peak remains acceptable. Checked (measured through tests) Thermocouple 2 Integrity Neither rupture nor movement of any parts of the item, or the item itself, should occur beyond a 1-meter range in any direction. If the integrity can be affected by the fire expos ure, tests under fire can be requested. To be checked under fire expos ure High speed Camera 3 Low noise No audible effect above 135 dB at one meter from the item in any direction. To be checked Sonometer 4 Arc and flames No electric arc, no flame should occur able to inflame material such as a sheet of paper whose density is 80 ± 10 g/m2 in contact with the item. To be checked High speed camera 5 Fumes No producti on of fumes, emanations or dust should occur which would reduce the visibility by 50% in a 1meter -cube chamber. The luxmeter is located at 1 meter from a constant light. To be checked Luxmeter Others 6 Inviolability To make the item safe to manipulate even in case of untimely firing, the design shall ensure that the pyrotechnic initiators can not be unscrewed from the separation nut, as a warrant of inviolability and safety for operators and users. Compliance of design options to be confirmed by regulation authorities Conception requirement Note 1: If no fumes are observed during tests performed for previous criteria, criterion 5) can be exempted. Note 2: The regulation authorit ies can determine that a packaged item is more dang erous than an unpackaged one and can request tests with packaging. Pyroalliance is in the process of verifying all of those 6 criteria until mid-2020 in order to submit the file and request the EU out of class certification. Finally, Pyroalliance objective will be to transpose the EU out of class certification to US regulation of goods by contacting US DoT. Conclusion While preserving all assets of the heritage design of its separation nuts (very high reliability and extrem ely fast actuation time), Pyroalliance is cur rently working and quickly progressing on bringing new advantages to its products: • With the addition of a new damping system, Pyroalliance has already reduced the level of induced shock by a factor 4 compared to its heritage design. As a next step, Pyroalliance is now targeting to reach a shock below 500 g SRS @2000Hz measured on a small spacecraft configuration • Besides Pyroalliance aims at delivering separation nuts out of the pyrotechnic classification (EU / US regulation law s) by mid -2020.

Document

AMS_2020.pdf

211 Lubricant Degradation in High- Load, High- Cycle Actuator Test Using Heritage Harmonic Drives for the Multi -Angle Imager for Aerosols Instrument M. Michelle Easter* Abstract The Multi -Angle Imager for Aerosols (MAIA) instrument features a pushbroom spectropolarmetric camera on a two- axis gimbal, which is actuated throughout the mission duration for multi -angle imaging, target revisiting, and inflight calibration. Each gimbal axis is driven by an existing design Mini Dual Drive Actuator (MDD A), which featur es a duplex brushless DC motor and redundant size 10 pancake Harmonic Drive (HD) gears et within its drive outputs. As the size 10 pancake assembly is no longer available from Harmonic Drive , LLC , units were acquired from long- term flight st orage. Due to the load and speed profiles, as well as the continuous operation required of the actuators, an endurance test was conducted to verify acceptability of the mechanism for the application, based on fatigue life. Approaching the two- life mileston e, the test failed due to breakdown of the Braycote 601EF lubricant, with a final apparent overall gearbox efficiency of 7%. Disassembly and SEM inspection revealed two distinct flavors of degraded and polymerized lubricant, tooth and race wear, material adhesion onto rolling wave generator bearing surfaces, and non- trivial amounts of corrosion. Due to the dither behavior and relatively high loads, it seemed breakdown due to high contact stress was the likely culprit. This led to the decision to change the drive output subassembly lubricant from Braycote 601EF grease and Bray Oil 815Z, to Rheolube 2000 grease and Nye Synthetic Oil 2001. Since the motors were already built using Braycote, and the lot of Harmonic Drives (HD) and duplex bearings had already been processed with Braycote, testing to verify compatibility and re- wettability of the two lubricants began. Meanwhile, a complete disassembly and inspection was performed, down to the individual bearingball level. The inspection revealed corrosion withi n the Wave Generator bearings, even on inner diameters of inner races that were bonded onto the Wave Generator hubs, despite having been stored in hydrocarbon oil. Specialized inspection also brought the black oxide coating application of the HD circular s plines into the spotlight, just as chemical analysis revealed an unexpected presence of copper in the degraded lubricant. Black oxide, old adhesive, and corrosion were removed from the HD assemblies, they were processed with Pennzane- based lubricant, and t he endurance test began again. This paper provides recommendations regarding use of mechanisms from long term storage, even if stored in seemingly ideal conditions, and document s a discovery process unique to work with heritage hardware. Also addressed is a performance comparison between Braycote 601EF and Rheolube 2000, as well as the Pin on Disc Test results regarding their compatibility along with evidence that, contrary to prevailing belief, it is possible to functionally rewet with Pennzane hardware th at was once processed with Braycote . Introduction MAIA Instrument The Multi -Angle Imager for Aerosols (MAIA) mission objectives are to assess the impact of mixtures of airborne particulate matter of different sizes and compositions on human health, and to collect multi -angle spectropolarmetric imagery over targets of interest with respect to air quality and climate research. The MAIA instrument is class 3, and features a pushbroom spectropolarmetric camera on a two- axis gimbal, which is actuated througho ut the mission for multi -angle imaging, target revisiting, and inflight calibration. * Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA Proceedings of the 45th Aerospace Mechanisms Symposium, NASA J ohnson Space Center, 2020

Document

AMS_2020.pdf

212 There are several heritage aspects of MAIA which are reminiscent of the Multi -angle Imaging SpectroRadiometer (MISR) mission, which has been using a multi - angle imaging approach to image aerosols in the atmosphere, imaging targets from nine angles, using nine cameras, since the year 2000. The reduction of the number of cameras required from nine for MISR to one for MAIA, is largely thanks to the Bi -Axis Gimbal Assembly (BGA), which articulates the MAIA camera to achieve multi -angle pointing. A key heritage item in the MAIA instrument design is the Mini Dual Drive Actuator (MDDA), which is a single- fault tolerant electromechanical device, used by MISR as its Cover Actuator, and as the North and South Calibration Plate Actuators. It also flew on Cloudsat and Galaxy Evolution Explorer (GALEX). The MDDA is valued largely due to its redundancy, providing the high- reliability desired to produce necessary mission -critical motions. The single- string Mini Uni -Drive Actuator was used on Mars Pathfinder for the High Gain Antenna Gimbal Actuators. Its design was inspired by the Standard Dual Drive Actuator, originally designed at JPL by Doug Packard. [ 1] The MAIA MDDAs will marry the applications of the Mars Pathfinder gimbal operation and the multi -angle imaging of MI SR, with one MDDA driving each of the two axes of the MAIA BGA. Figure 1. Mini Dual Drive Actuator However, the MAIA project seemed to be in luck, as there existed a set of spare Harmonic Drive units in flight storage, which had been originally purchased by the Cloudsat project. The Cloudsat actuator build had occurred just following MISR in the 1990s, and used the MDDA as its Reflector Mechanism. As fate would have it, the lot of flight spare size 10 pancake HD s had been preserved in their original heat -sealed packaging, l ocked away in flight stores since their receipt in 1995, as delivered by their namesake manufacturer who, at that time, was operating under a different ownership than the Harmonic Drive LLC of today. Flight spare duplex output bearings were also identified, as well as spare Duplex Motors, which would end up getting sent back to Ducommun Technologies , their original manufacturer, to be rewound and have bonds, wires, and Hall sensors replaced, for the new MAIA application. MAIA Actuation and the Endurance Test In flight, the two MAIA MDDAs will drive two slightly different load and speed profiles, with sweeps as well as stop and stare motions, dependent on the axis for which they are designated and the mode of operation of the instrument. The nominal expected output loads are around half of the maximum load rating of the size 10 HD, which is documented at 5.1 Nm ( 45 in- lb), and the application was one of continuous use throughout the mission life. Even with heritage and redundancy, no other MDDA application was similar enough to avoid some testing. T his called for the execution of a new test, with the intention of verifying the fatigue life of the mechanism and its endurance with the MAIA operational profiles. For simplicity, a hybridized profile was created, to capture both the higher loads and the requirement for repeated gear tooth engagement of the off-track axis profile, and the higher speed and larger rotational

Document

AMS_2020.pdf

213 output distance required of the on-track axis. The profile was conservative, and yielded a series of test events to be repeated in cycles, each with a 2.3 Nm ( 20 in -lb) continuous applied output load being driven at a speed of eight degrees per second, with dither frequencies requiring direction changes every 1.5, 6.5, or 16 seconds. The test would run for three times longer than one mechanical life, with a total runtime of almost 700 hours, equating to approximately 13.5 million equivalent Wave Generator input revolutions. Additionally, since the updated MAIA Du plex motors were specified to have a minimum stall torque of 184 mNm ( 26 in- oz), and the HD gear ratio is 244:1, a risk to be addressed was that the actuator could easily exceed the 5.1 Nm ( 45 in- lb) momentary peak (ratchet) torque rating of the size 10 HDs. Although test records could not be located, the memory remained of the earlier completion of static, sustained noratchet verification testing up to a 11.3 Nm ( 100 in- lb) output load, completed for the MISR program. Thus, passing this test before, after , and throughout the three times life program would also be a success criteria for the Endurance Test campaign at large. The element of the MDDA most susceptible to wear due to fatigue is the HD output gearing. One of the two redundant drive output subass emblies served as the test article, shown in Figure 2. The HD was to be grease plated using Braycote 601EF and the duplex bearings with Braycote 600EF, both using Brayco 815Z oil, reflecting the heritage grease plating and packing method. The unit was mounted into a simple dynamometer fixture, driven by a brushed test motor, loaded with a brake, along with an input torque transducer and output transducer measuring torque and speed. The simplicity of the setup was intended to allow for continuous operation, to be monitored by a chart recorder, and checked on periodically by engineering. Figure 2. Endurance Test Article, the MDDA Drive Output Subassembly The HDs, and a spare lot of the necessary 440C duplex output bearings, were pulled from flight storage and taken to the cleanroom for cleaning and inspection. The Wave Generator subassemblies had been delivered with their AISI 52100 bearings and 304 CRES spacers bonded onto the Wave Generators hubs, and, per the original source control drawing, a black oxide coating had been applied to the ductile iron circular splines and the AISI 1144 Stressproof Steel® Wave Generators per MIL- DTL-13924. Everything was heavily coated with a layer of a lightweight water -displacing hydrocarbon oil. Similar to the HDs, the Timken duplex bearings were still sealed in their original packaging, also covered in hydrocarbon oil. The HDs were partly disassembled, leaving the bonded Wave Generator subassemblies intact, and ultrasonically cleaned using acetone, followed by Vertrel XF , and a final oven bake. They were inspected under 15x magnification by engineering and quality and, after receiving passing inspection results, the units were immediately grease plated using Braycote 601EF (10% by mass) and Vertrel (90% by mass), reinspected, and sealed in moisture barrier bags under dry nitrogen gas purge. The same process was followed for inspecting and processing the duplex bearings, using Braycote 600EF, and all of the packages were stored in the cleanroom to await further assembly st eps. When the Drive Output subassembly was completed, a layer of Braycote 601EF was applied to each of the Wave Generator bearings as required to obtain 100% ball coverage, which was manually run- in while also adding 12 drops of Brayco 815Z Oil to

Document

AMS_2020.pdf

214 each bearing (each of which also contains a phenolic retainer). Braycote 601EF was manually added onto all circular and Flexspline teeth, with excess removed after assembly. Endurance Failure The First Life The test began with completion of start -up torque measurements of the Drive Output subassembly, and an average value of 11.1 mNm (1.57 in- oz) was observed, landing nicely within the acceptance criteria of <14.1 mNm ( 2.0 in -oz), per the original size 10 HD source control drawing. The required sustained static output load of 11.3 Nm ( 100 in- lb) was maintained for 60 seconds with no ratcheting behavior observed, in both directions. The subsequent running portion of the test kicked off, and for the first 27 hours, the input torque required to drive th e output torque load of 2.3 Nm (20 in- lb), alternating directions every 6.5 seconds at a rate of 8.0 deg/s , was an average of 38.1 mNm ( 5.4 in- oz). The input torque dropped to 35.7 mNm (5.05 in- oz) (-6.5%), shown in the left panel of Figure 3, corresponding with an increase in speed of 0.58 deg/s (+6.7%) across a 52 minute period. The speed returned to its approximate initial condition (there was no speed controller in the configuration) starting around ~170 hours, a nd the input torque increased, to around 41.7 mNm ( 5.9 in- oz). Then, in the final 16.5 second dither time test event , at around 210 hours, the output speed dropped to about 87% of its starting value, and the input torque increased to a maximum of 45.2 mNm (6.4 in- oz), about 18% higher than the starting value. However, after about 30 hours, the speed and input torque seemed to level back out, ending with an average input torque of ~ 38.1 mNm ( 5.4 in- oz) and an output rate of about 8 .0 deg/s. The first life w as complete, reaching over 250 hours of as -run test time. Despite the variation along the way, the consistent starting and ending input torque and speed measurements seemed like indicators of a successful first life. The startup torque was re -measured, and an average value of 4.8 mNm ( 0.68 in- oz), still defined as passing, was observed . The sustained 11.3 Nm ( 100 in- lb) no-ratchet test was successful again, and life two was ahead. Figure 3. Comparison of input torque observed over the initial 45 hours of the first and second lives. The Death March At the onset of life two, the average input torque was back down to 35.7 mNm (5.06 in- oz), which corresponded nicely to the run- in value seen after the first life settled. As shown in Figure 3 and, although the speed was around 12% higher than it was at the close of the previous test, the lower measured torque reduced concern associated with the higher speed, and the test proceeded. Both measured values climbed slowly to 38.1 mNm ( 5.4 in- oz) and 9.0 deg/s over the course of the following 75 hours. Then, another change appeared with about a 0.5 deg/s speed drop over a half hour period, accompanied by an input torque increase to 41 mNm ( 5.8 in- oz).

Document

AMS_2020.pdf

215 This marked the beginning of the roller coaster ride to c ome. Speed increased, then decreased, then increased again over the following 15 hours, reaching a local high of 9.77 deg/s. After another 10 hours, the speed had dropped back down to 8.4 deg/s. The input torque followed the pattern of increasing with the speed decreases, and vice versa, over this 56.4 hour run with a 16 second direction change timing. At the profile end, input torque was an average of 43.4 mNm (6.15 in- oz). After this, the death march began. At the start of each event type to follow, as well as in between, the motor voltage was readjusted to obtain the desired speed, and the input torque would swell in response and subside again as the speed would drop back down, like a snake swallowing prey (or a system without speed control). The effect of one such adjustment can be seen in the middle of the 1.5 second dither time test event in the shaded center panel of Figure 4. After 34 hours, the sy stem stop ped cycling , and the measured input torque seemed to change exclusively in the uphill direction, until it reached an average of almost 72 mNm ( 10.2 in- oz), or double the observed value at the start of life two. See Figure 4 and Figure 5. Figure 4. Output Speed Variation Figure 5. Input Torque Variation Observed During the Second Life Observed During the Second Life In the final stretch of life two, a marathon run with a 6.5- second dither time delay was embarked upon. The speed had dropped from the initially set value, but seemed to settle out at just over 7.0 deg/s, which lasted almost 15 hours. Then, things declined. Over a three hour period, the speed dropped from 7.0 to 4.5 deg/s, which was the lowest yet observed speed, shown in Figure 6, and the input torque increased to the highest yet seen value of 91.1 mNm ( 12.9 in- oz), shown in Figure 7. The passing of 5 more hours brought hope, then another 5 hours brought despair. When it seemed like the system had settled itself at 4.5 deg/s and a drag torque of just over 84.7 mNm ( 12 in- oz), the motor voltage was driven back up to obtain the desired output speed. Figure 6. Output Speed Degradation i n Figure 7. Input Torque Increase in the the Final 6.5 Second Dither Time Test Event F inal 6.5 S econd Dther Time Test Event This time, when the speed was increased to the desired value, the system faught harder than before. Input torque climbed quickly , exceed ing previous test maximums, as seen in the far right of Figures 6 & 7. In

Document

AMS_2020.pdf

216 addition to the behavior was a simultaneous temperature spike of 10 °C on the test article housing in under a half hour, with no change to the ambient environment. When the data for both lives was concatenated, as shown in Figure 8, it was clear that this final 6.5 second dither time test event showed unacceptable performance. The test was halted for further assessment. Figure 8. From Top to Bottom, Measured Motor Cur rent, Speed, Housing Temperature, Input Torque and Output Torque from the Endurance Test, Throught the Final 6.5 Second Dither Time Test Event The Final Surrender For the endurance test that had been completed, data had been logged at 1 Hz. It was decided to perform a final test run with a higher sample frequency, holding the output speed to the 8.0 deg/s target value, and running with a 16- second dither time delay under the full 2.3 Nm (20 in- lb) load. This one last batch of data was desired before the unit was to be disassembled, never to be run in the same configuration again. This final run followed the pattern of the input torque and speed rising and falling in terrifying opposition, growing increasingly worse as time processed. The test reached a climax when the input torque values climbed to a whopping 141 mNm (20 in- oz), where it sat for about 26 hours, with the speed continuing to descend throughout. Then, the system turned a strange corner. The input torque dropped, about 40 minutes before which t he titanium housing temperature climbed 5 degrees Celsius in about 20 minutes. The output speed experienced a major noise reduction at the onset of this temperature increase, and proceeded to exceed 10 deg/s, the highest value observed yet. A few blips fol lowed, before the input torque necked down to 77.7 mNm (11 in- oz). See Figure 9. After seven hours, the rollercoaster ride ended, as the motor began to stall. See Figure 10 for the full test dataset. The time had definitely come for disassembly. Beforehand, the start-up torque and no- ratchet torque tests were conducted. The no- ratchet torque was successful again, and the start -up torque test pushed the acceptable limits, with a peak measured value of 24.7 mNm (3.5 in-oz).

Document

AMS_2020.pdf

217 Figure 9. From Top to Bottom, Speed, Input Torque, and Temperature up to Termination of the Final 16 Second Dither Time Test Event Figure 10. From top to bottom, speed, input torque, and housing temperature for entire test, including finally executed 16s dither time added test event. Vertical lines indicate test event separation points. Sawteeth in temperature data induced by local thermostat. Removal of a single piece part revealed severely degraded lubricant in the Drive Output subassembly. The grease within the Wave Generator bearings looked like copper colored glitter, and felt like the glitter was mixed with dried glue. The material surrounding the tooth interface of the Flexspline and the Circular Splines was black in color and crumbly, with a texture similar to clay mixed with brown sugar. Refer to Figure 11, panels 1 and 2 show the blackened sample within the gear mesh, and panels 3 and 4 show the glittery Wave Generator bearing samples. Each sample was sent to the analytical chemistry lab for analysis, and the HD assembly was ultrasonically cleaned using Vertrel XF, then sent for Scanning Electron Microscope (SEM) imaging and Energy Dispersive X -Ray (EDX) analysis .

Document

AMS_2020.pdf

218 Figure 11. Panel 1: Drive Output Subassembly After Removal of Dynamic Spline, Showing Output Tetth of Flexspline, and Static Spline. Panel 2: Dark, Crumbly Material Gathered in the Static Spline Feature During Disassembly. Panel 3: Wave Generator Subassembly with Flexspline Installed. Panel 4: Opened Drive Output Subassembly Before Distrubance of HD Components The HD showed wear on all teeth, biased with the worst damage on the dynamic spline tooth interface, and on most Wave Generator bearing surfaces. Damage on the teeth and bearing races was aligned with the line of actio n of the bearings. SEM revealed wear to the Flexspline teeth, shown in Panel 4 of Figure 12, which corresponded with Dynamic Circular Spline tooth wear, shown in Panel 1 of Figure 12. The Flexspline had been slipping on the wave generator bearing outer rac es, with polymerized grease product in between, and as a result, damage was seen on the inner diameter of the Flexspline (shown in Panel 2 of Figure 12) and the outer diameters of the bearing outer races (shown in Panel 5 of Figure 12). Flexspline material was removed at that interface, combined with the bearing grease product, and redeposited onto the Wave Generator bearing balls (see Panel 3 of Figure 12) and ball tracks. Figure 12. Panel 1: Dynamic Spline tooth wear. Panel 2: Damaged Flexspline, input side inner diameter - damage pattern reflects ball track load path. Panel 3: Flexspline material adhesion onto bearing ball (increased chromium & nickel seen as dark patches via EDX). Inset shows representative “normal” ball surface at similar scale. Panel 4: Flexspline output side tooth wear. Panel 5: Damaged bearing outer race- pattern reflects ball track load path.

Document

AMS_2020.pdf

219 Raman spectroscopy of the samples revealed the coppery hue to be due to iron oxide, and EDX SEM results showed increased concentrations of oxygen on the bearing races and ball surfaces. Both results indicated the presence of corrosion. X -Ray Fluorescence (XRF) detected the presence of iron, chromium, nickel, and copper in both grease samples. Iron, the Circular Spline material, and chromium and nickel, constituent materials of the Flexspline, corresponded with the worn teeth observed on both via the SEM inspection. The presence of copper was an unexpected discovery. Table 1. Anion Content for Harmonic Drive Braycote Samples, Expressed as M icrogram Anion in the Water Extract used in S onication, per gram of Braycote (ppm). Sample Amount (mg) Fluoride (ppm) Chloride (ppm) Control – New Braycote 601EF 27.51 < 0.88 4.60 Braycote 601EF – Bearing Sample 26.54 45.5 < 2.3 Braycote 601EF – Tooth Sample 25.76 110 3.38 Anion levels measured using Ion Chromatography in the blackened tooth grease sample and the glittery bearing grease sample are shown in Table 1. Extremely high fluoride levels in both samples, especially in the tooth sample, indicat ed Braycote degradation had occurred. The higher Fluoride level in the tooth sample seemed due to the higher stress experienced by the grease at that interface, due to the sliding interaction between the HD spline teeth. The Braycote had clearly polymeriz ed and the grease performance and appearance, with the wear patterns along the ball paths of the splines and bearings, were reminiscent of failures described by Conley and Bohner in their report on early perfluoropolyalkylether (PFPE) lubricant testing. [ 2] It seemed the continuous operation of the HD at a load which was 45% of its maximum rated value, along with the dither -like motion profile, were too much for the Braycote to handle. The severe degradation of the Braycote, especially in the high-shear HD sliding gear mesh, and the presence of the notorious Lewis acids in the samples (higher at the tooth interface) seemed consistent with the summary of PFPE breakdown under high stress by Herman and Davis. [ 3] The conclusion was drawn that breakdown of Brayc ote occurred due to the high loads and cycles, and polymerization of the grease caused lubricant starvation of the assembly, increasing the rate of wear to the splines and bearing components. The rollercoaster effect apparent in the input torque and output speed test data appeared to demonstrate the effect of the generation of the friction polymer and its reaction with oxides to produce metallic fluorides to act as in situ solid lubricants (yielding a momentary improvement in performance), followed by the effect of these fluorides as catalysts to accelerate the PFPE breakdown (resulting in subsequent drops in performance). [ 3] Path to Resolution Due to the status of the MAIA program, it was too late for load reductions, and the MDDA design similarly could not be changed. The decision was made to change lubricants for the Drive Output Subassembly from PFPE Braycote (Braycote 601EF and Brayco 815Z) to hydrocarbon- based Pennzane products (Rheolube 2000 and Nye Synthetic Oil 2001) to leverage its increased lifetime capability. While known to demonstrate an increased lifetime and a decreased wear rate, use of Pennzane lubricants is also known to reduce cold temperature performance, due to its increase in viscosity with reducing temperature. [ 4] The in- flight thermal environment for the MDDA would be benign, operating nominally somewhere between 0°C to 30 °C, so obtaining adequate performance in flight was not at risk, but demonstrating torque margin at the more extreme cold temperatures required in ground testing was. Additionally, no experience could be referenced using Pennzane with ductile nodular cast iron for flight, so material compatibility needed to be established. Furthermore, the Duplex Motors used for the MDDAs had

Document

AMS_2020.pdf

220 already been built and delivered with Braycote 600EF, and, since no appropriate seals existed in the MDDA design, a cross -contamination question was raised regarding the potential impact of migration of one lubricant into the region of the actuator where the other was located. Finally, there was extreme uncertainty whether the part lubricated with Bray could be reliably relubricated with Pennzane due to differences in wetting characteristics. Three tests were executed to provide confidence in: (a) the compatibility of the Pe nnzane with the HD component materials, (b) the ability to maintain MDDA performance in the event of cross -contamination due to lubricant migration, and (c) the ability of the MDDA to achieve 100% torque margin against a 5.1 Nm (45 in- lb) requirement at th e MAIA coldest operational test temperature of -20°C. For (a), an accelerated test would be completed, running the Drive Output Subassembly with the updated Pennzane lubricants, after which inspection and chemical grease analysis would be completed. Regarding item (b), a Pin- OnDisc test was to be completed using a mixture of Rheolube 2000 and Braycote 600EF. Finally to address (c), cold temperature testing would be completed to evaluate the trade between maximizing the Pennzane grease and oil fill amounts and minimizing the cold temperature input drag. Rewetting & Initial Accelerated Test The entire lot of HD units and Duplex bearings had been cleaned, inspected, and processed with Braycote lubricants together, and thus, it had to be removed from all of them, including their phenolic retainers. The units were all sent to the Laboratory for Applied Tribology at JPL, to be cleaned and re- processed. To remove the Braycote, the parts were placed into a vapor degreaser, where they sat for several days, followed by ultrasonic agitation, both using Vertrel XF. The Wave Generator subassemblies still remained intact, with bearings and spacers bonded onto the Wave Generators. Their disassembly was discussed, but concerns that were presented regarding potential damage to the irreplaceable flexible bearings during the disassembly and reassembly process prevailed. Some patches of corrosion were visible on the Wave Generator bearing races, which were removed using Scotchbrite. The HD and the duplex bearings were then grease plated using a 10% Rheolube 2000 and 90% Heptane by mass solution. An additional gre ase fill of Rheolube 2000 with a mass of 0.1151 g was added to the Wave Generator bearings, resulting in 0.0027 g of Rheolube mass applied per each of the 42 bearing balls shared across the pair of bearings. In this configuration, the HD was reassembled into the Drive Output Subassembly along with a similarly prepared Duplex bearing pair, and the accelerated test was conducted using the Endurance Test fixture. The test article drove a 3.4 Nm ( 30 in- lb) load in the clockwise direction for 72 hours consecut ively at a rate of 16 deg/s. No significant change resulted in input torque nor in speed across this test. The near -constant input torque measured across the test gave confidence in the ability to re- wet the assembly with Pennzane and to subsequently perform well under an aggressive set of initial conditions. After completion of the 72 hours, the test article was disassembled, revealing the Pennzane inside to have a normal consistency, but to be grey in color. The grease sample was sent to the analytical chemistry lab to look for any signs of a potential impending lubricant breakdown. The gray color of the lubricant was verified to be primarily iron and chromium, with a small amount of titanium (the housing material), by XRF imaging. Fourier Transform Infr ared spectroscopy (FTIR) was performed to determine that no significant changes had occurred in the bulk grease chemistry, and there was no evidence of oxidation in the sample. The Direct Analysis in Real Time - Mass Spectrometry (DART - MS) method showed no significant change in the molecular weight of the sample when compared to a comparable new, un- tested sample as well. Overall, the accelerated test yielded a passing result, verifying that the HD and the duplex bearing pair had been successfully rewetted, and the Pennzane lubricants could be trusted to be materially compatible with the HD. Pennzane Cold Testing The components were cleaned using Heptane to remove the Pennzane, and re- coated with a fresh Rheolube 2000/Vertrel XF 10/90 mass ratio grease plat e layer. This time a Nye Synthetic Oil 2001/ Rheolube 2000 slurry with a 50/50 mass ratio was also applied to the Wave Generator bearings and a 70/30 oil/grease ratio applied to the duplex bearings. The duplex bearings received 0.077 g of slurry to each

Document

AMS_2020.pdf

221 of the two bearings. A total of 0.11 g of the 50/50 mass ratio slurry was added to each Wave Generator bearing. The HD and duplex bearings were then reinstalled into the Drive Output Subassembly, which was installed into the Endurance Test fixture. The syst em was run with no applied load at speeds of 8.0 and 4.5 deg/s and temperatures of -20°C, -8°C, and 30°C. Startup torque tests were also conducted over temperature. The results of this testing are shown in the panels indicated as “Cold Test 1”, of Figure 13. Another test was run with updated ratios, to understand the effect of reducing the oil percentage while increasing overall mass of applied slurry. The Nye Synthetic Oil 2001/Rheolube 2000 ratio this time was 39/61 by mass for the Wave Generator Bearin gs, and 36/64 for the duplex bearings. The resultant drag and startup torque values obtained over temperature are shown in the panels indicated as “Cold Test 2”, of Figure 13. See Table 2 for lubricant parameters related to both Cold Tests. Figure 13. Test Data from the C old Tests Conducted to D efine Grease & Oil Fill Quantities. The second tested lubricant configuration was chosen due to the improved performance, combined with the additional slurry mass of the assembly. The desire for a higher slurry mass was to minimize any future risk of lubricant starvation. The drag torque values were acceptable, thanks to the high- torque output of the re-wound Duplex motors, and the relatively high current available from the instrument power bus. Table 2. Lubricant Application Parameters for Cold Tests 1 & 2. Cold Test Item Nye Synthetic Oil 2001/Rheolube 2000 Ratio (% mass) Slurry Mass per Bearing (g) 1 Wave Generator Bearings 50/50 0.11 1 Duplex Output Bearings 70/30 0.077 2 Wave Generator Bearings 39/61 0.137 2 Duplex Output Bearings 36/64 0.115

Document

AMS_2020.pdf

222 Pin-On-Disc Testing To address the concern of cross -contamination damage due to potential inter -mixing of Pennzane and Braycote, Pin- On-Disc testing was conducted with a 50/50 by mass mixture of Braycote 600EF and Rheolube 2000. The test was conducted at a 0.77 m/s rate in one rotational direction, with a pin pressure of 500 MPa. The result is shown in Figure 14. An increase in the coefficient of friction (CoF) was seen at the test onset, which settled relativel y quickly after about 200 m, staying constant until around 3.5 km. After this, the measurement noise and the CoF both increased and fluctuated slightly until the test was completed after 7.5 km of pin travel through the lubricant mixture. The test demonstr ated that there would be no significant detriment to performance with a potential long- term mixing of the Rheolube and the Braycote. A Final Lubrication with an Unexpected Twist The results of the Pennzane accelerated test, cold testing, and Pin- On Disc testing all gave the green light to reattempt the Endurance Test with the new lubricant configuration. A final spare, un- used and undesignated HD remained, and was selected to be the first to receive the new treatment. The unit was cleaned first with the vapor degreaser and Vertrel XF to remove the Braycote 601EF, then reprocessed using a 10/90 by mass Rheolube 2000/Heptane grease plate, followed by a 40/60 by mass Nye Synthetic Oil 2001/Rheolube 2000 slurry fill process. When the HD was installed into the Drive Output Subassembly, a small piece of debris was found in the Wave Generator bearing lubricant, accompanied by a crunchy feeling upon bearing rotation. The unit was re- cleaned, re- inspected, and received an additional abrasive touch- up to remove some newly found corrosion. New lubricant was applied, and a similar result occurred with a mysterious grease discoloration discovered upon higher -level assembly of the HD, accompanied by more bearing crunch. Figure 14. Pin-On-Disc Results for 50/50 by Mass Braycote 600EF/Rheolube 2000 L ubricant on 440C (58-62 HRC) S ubstrate in a Uni-Directional Test Configuration with a P in Pressure of 500 MPa and R ate of 0.77 m/s. Thus the time had come to completely disassemble the Wave Generator subassemblies, down to the individual ball level, for inspection. Each bearing was pressed off of the Wave Generator, preserving serialization between the two sets of bearing components. T his revealed a significant amount of corrosion, on the bearing balls and races, and along ball paths. The source of the crunchiness was confirmed. The corroded surfaces were abraded as much as possible, the components were cleaned with boiling deionized water followed by Heptane, reassembled into bearings, which were re- bonded onto the Wave Generators. The Wave Generator assemblies were then re- grease plated and filled, as before. The coppery appearance of the bearing sample had been due to high levels of iron oxide, according to the Raman spectroscopy results. It was highly likely that similar high levels of corrosion hiding in the corners of the Wave Generator bearings in the previously failed Endurance Test helped accelerate the PFPE degradation, readil y able to help produce the high Lewis acid fluoride contents seen in Table 1.

Document

AMS_2020.pdf

223 All the while, a simultaneous deeper inspection of the Circular Splines revealed that their black oxide coatings did not demonstrate the expected physical properties. Namely, it seemed unnaturally thick for a hot black oxide conversion coating, and instead could be scraped off with a dental pick, turning to powder in the process and revealing a bare surface. It appeared to be a room- temperature, or cold, black oxide application, which is a deposited layer of copper selenium as opposed to a conversion coating, which is not included in the MIL- C-13924 specification as an acceptable black oxide application process. This understanding of the black oxide type provided the clue to the rogue copper discovered in the previous XRF grease analysis results. The cold temperature black oxide coating had been slowly scraped into the grease by the sliding tooth action during the Endurance Test, and, in addition to generating debris in the tooth mesh grease, it was certainly not performing as a corrosion inhibitor for the Circular Splines, as it was intended. This meant the Circular Splines had to be abraded as well, to remove all of the undesirable black oxide coating. After this was completed, the units were re- cleaned with Heptane, re- grease plated, and slurry filled as before. This time, when the completely renovated HD was reassembled into the Drive Output subassembly, rotation was smooth, and lubricant was clean. See Figure 15 for Circular Spline images before, during, and after cleaning and re- assembly. Figure 15. Panel 1: Harmonic Drive with Room Temperature Black Oxide on Circular Splines, Dynamic Spline Shown, Installed into Drive Output Subassembly, with Braycote 601EF at Tooth Interface. Panel 2: Clean Circular Splines with Black Oxide R emoved. Panel 3: Clean Harmonic Drive Circular Spline, Dynamic Spline shown, Installed into Drive Output Subassembly, with Rheolube 2000 at T ooth Interface. The Final Endurance Test The new Pennzane lubricant configuration needed to be vetted by running it through a successful Endurance Test. A refined test profile was created, reducing the loads and varying them based on the three dither timings. The target speed was reduced to be the flight maximum value of 7.75 deg/s, instead of 8.0 deg/s. Cycle counts were scrutinized, and the definition of an actuator mechanical life was reduced from 4.5 to 3.7 million input revolutions. The test would be run for a total of two time one life instead of three times , reflecting a hesitation against over -conservatism, since the JPL standard policy calls for a 2x mechanical life test for wear -life limited elements. This brought the required total HD input revolutions for a successful test declaration down from 13.5 to 7.4 million. The fixture was also modified to include a pulley and weights at the output, such that the brake would be used to apply different bi -directional output drag loads, and the pulley system would simulate the uni -directional load of the flight ant i-backlash mechanism, present in the BGA design to facilitate pointing accuracy. A simple control system was also implemented, with a speed controller and position- based sensor feedback to make the testing automatic and consistent.

Document

AMS_2020.pdf

224 Figure 16. Harmonic D rive Lifting Efficiency vs. Time during the Final Endurance Test Initial startup torque values were low with an average measured value of 6.0 mNm (0.86 in- oz), compared to the 11.1 mNm (1.57 in- oz) for the Braycote 601EF initial measurement. Variations seen in input torque and output speed across this test series were due to operator error in tuning of the bi -directional applied load and speed setting. Despite some setting errors, the telltale sign of the success of the Pennzane Endurance Test was the slight but steady increase in the HD lifting efficiency over the total 400 hour period. Corresponding with the increase in efficiency over the two times one mechanical life, the final measured startup torque after test completion had actually reduced from 6.0 mNm (0.86 in- oz) to 5.1 mNm (0.72 in-oz). The test was finally over. Conclusions The long standing controversial relationship between Pennzane and Braycote was challenged by the MAIA MDDA application. The Pennzane ability to functionally rewet hardware t hat had previously been processed with PFPE lubricant was demonstrated by the completion of the accelerated 72 hour test. The test showed no change in the efficiency over its duration, despite the aggressive load and speed profile. This indicated that the Pennzane effectively lubricated the rolling and sliding surfaces of the HD and Duplex bearings, and maintained this effectivity for 72 consecutive hours in one direction. Post -test chemical analysis showed no indications of any impact to the lubricant itself due to the operational profile. Thermal testing also showed the re- wettability to persist over temperature. The Pin- On-Disc test results gave confidence regarding the chemical compatibility of the lubricants over a long life, and their ability to lubric ate effectively despite their intermixing. These test results gave the MAIA project the confidence to proceed with the unconventional solution of changing to Pennzane- based lubricants in the middle of the program, resulting in an MDDA assembly that has a B raycote lubricated motor, and rewetted Pennzane output gearing, all in one package. The relatively uncommon process of pulling flight hardware from storage to revamp for a new flight application provided challenges to the team. The long- term storage conditions, bagged and sealed in hydrocarbon oil, proved to be inadequate for the 52100 HD Wave Generator bearings over the two decades the units were shelved. To reliably use this inherited hardware, a full disassembly, down to the individual bearing-ball level, and detailed chemical and abrasive cleaning was required in order to locate and eliminate all of the hidden locations of the corrosion that had developed over time. For certain components to be used from long- term storage, a similar total d isassembly and detailed inspection is recommended, even if the assemblies appear to be acceptable by external visual inspection. Specifically, those that use materials that are sensitive to corrosion, such as 52100 steel, and are inherently difficult to inspect due to hidden geometry, such as bearing assemblies. Trust, instead, that corrosion is sneaky and will wait hiding in a crevice or shadow, to ruin hardware performance when least expected. Finally, the black oxide surprise turned out to be another unexpected result of using hardware from long term storage. The HD units used for the MAIA project were originally received in the 1990s, at a time when

Document

AMS_2020.pdf

225 their manufacturer was not yet certified in compliance with ISO 9001. At the time, material certifications were not required as a part of purchase orders (which is a requirement now). A perfect storm of conditions set the stage for the application of the cold temperature black oxide to be missed, despite the long duration they were possessed. If the units had been part of a lot that was intended to be used in short order, the discovery would have been made during testing of the other projects. The use of hardware drives being purchased as a solitary lot, intended to be immediately shelved, and the minimal involvement from the quality organization at the time of purchase, meant the discrepancy snuck in, undetected, to be discovered instead during the development of the next generation of MDDAs decades later. As a result of this discovery, a recommendation is made for heightened inspection to be completed of hardware that was either received before quality process changes, or before any significant changes in the ownership of or certifications obtained by the supplier. References 1. Packard, Douglas T.; “Dual Drive Actuators”; Proceedings, 16th Aerospace Mechanisms Symposium; May 13- 14, 1982; NASA CP -2221. 2. Conley, Peter L. and Bohner, John J.; “Experience with Synthetic Fluorinated Fluid Lubricants”; Proceedings, 24th Aerospace Mechanisms Symposium; April 1 1990; N ASA- CP-3062. 3. Herman, Jason and Davis, Kiel; “Evaluation of Perflouropolyether Lubricant Lifetime in the High Stress and High Stress -Cycle Regime for Mars Applications”; Proceedings, 39th Aerospace Mechanisms Symposium; May 1, 2008; NASA/CP -2008- 215252. 4. Venier, Clifford; Casserly, Edward W.; Jones, William R., Jr.; Marchetti, Mario; Jansen, Mark J.; Predmore, Roamer E.; “Tribological Properties of a Pennzane- Based Liquid Lubricant (Disubstituted Alkylated Cyclopentane) for Low Temperature Space Applications”; Proceedings, 36th Aerospace Mechanisms Symposium; April 1, 2002; NASA/CP -2002- 211506. Acknowledgments This work was performed at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Sp ace Administration. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not constitute or imply its endorsement by the United States Government or the Jet Propulsion Laboratory, Pasadena, California. © 2020. California Institute of Technology. Government sponsorship acknowledged. The author would like to thank JPL chief engineer, and original delivering engineer of the MDDA, Michael R. Johnson, from the Jet Propulsion Laboratory (JPL), for patiently providing his never ending technical guidance and mentorship. Also, thanks to Allison Ayad, for her tireless support during the failure investigation and her valuable data reduction, also through JPL. Thank you to Duval Johnson, of the Laboratory for Applied Tribology at JPL, who spearheaded the disassembly, cleaning, and detailed inspection of the Wave Generator subassemblies, and skillfully completed the Braycote removal and rewetting of the Harmonic Drives and duplex bearings with P ennzane. Thank you to the expertise of Mark Balzer, also of the Laboratory for Applied Tribology at JPL, who identified the cold temperature black oxide coating. Many thanks also to the Analytical Chemistry Laboratory at JPL, namely Nick Heinz, Mark Anders on, and Bill Warner, who performed all chemical grease analysis described in this paper. Lastly, but not least, the author would like to make a special acknowledgement of Harmonic Drive, LLC. Harmonic Drive, under their current ownership, has successfully provided flight hardware to JPL for decades. In recent years, the author of this paper, specifically, has received more than thirty pancake style HD units (in size 14) which have been inspected and approved by JPL QA, and have been integrated into high reliability flight actuators for multiple projects, including successful life test completion. No anomalies related to any of the HD units have resulted during these program developments.

Document

AMS_2020.pdf

227 Multi -Mission Deployable Boom: Spring Mechanism Design, Failure Investigation, and Resolution Christine A. Gebara* and Paul D. Lytal * Abstract As more ambitious missions are being pursued to better understand Earth and our Solar System, aerospace mechan isms and deployable structures are being approached in new fas hions. NASA’s Jet Propulsion Laboratory is developing two low Earth orbit satellites that will use radar instruments to better understand temporal changes in the Earth’s surface . Both the Surface Water Ocean Topography (SWOT) and the NASA- ISRO Synthetic Aperture Radar (NISAR) spacecraft utilize large carbon fiber deployable antennae to conduct such science . These large antennae have been des igned with similar mechanisms. During testing of the spring mechanisms that deploy the antennae , a hardware failure was found. The source of the hardware failure was traced back to the custom torsion springs used within the mechanism. Because of the mechan ism volume constraints , the springs were designed with high aspect ratio rectangular cross sections to maximize the spring constant for the me chanism. Ultimately, a failure investigation and testing campaign led to spring mechanisms that have been successf ully integrated into both spacecraft . Introduction NASA’s Jet Propulsion Laboratory (JPL) is currently developing two earth orbiting satellites. The Surface Water Ocean Topography (SWOT) mission will conduct the first global survey of Earth’s surface water. The NASA- ISRO Synthetic Aperture Radar (NISAR) mission will study temporal changes to Earth’s land and ice-sheets using advanced radar techniques. Both missions serve to better understand how the Earth is changing over time using radar -based instruments. The SWOT mission is a collaboration with the French, Canadian, and UK Government Space Agencies and will launch from Vandenberg Air Force Base on a SpaceX Falcon 9 Rocket in 2021. NISAR is a partnership between NASA and the Indian Space Research Organization (ISRO) and will launch in 2022 from Satish Dhawan Space Center on an Indian Geosynchronous Satellite Launch Vehicle. Figure 1. A) SWOT with hinge locations shown on one mast B) NISAR with hinge locations shown. Both SWOT and NISAR use deployable radar reflector mast designs developed at JPL. These deployable masts, while different in geometry, have similar components and sub- assemblies. Both masts are constructed from bonded Invar and carbon fiber composite structures and employ analogous flight * NASA Jet Propulsion Laboratory, Pasadena, CA Christine.A.Gebara@JPL.NASA.gov ; Paul.D.Lytal@JPL.NASA.gov Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

228 deployable hinge mechanisms. The SWOT mission has two identical reflector masts, each with two deployable hinges. NISAR has a single mast with four deployable hinges. These masts can be seen in Figure 1. Operationally, the deployable masts are launched in a stowed state wi th a launch restraint system composed of separation nut devices. When commanded, the launch restraints release a pre- tensioned spring and damper mechanism which deploys each hinge. Hinge deployment progress is monitored on the ground using a potentiometer as well as a limit switch on each hinge. Upon completion of the deployment, an actuator -driven latching mechanism preloads precision alignment features on either side of the hinge together. Figure 2 displays a n overview of the mechanisms. Figure 2. Hinge Deploy & Latching Mechanisms (NISAR configuration s hown) Mechanism Design and Fabrication Mechanism Design Each deployable hinge for the SWOT and NISAR masts is outfitted with a spring, damper and potentiometer mounted co-axially with each hinge line. The NISAR mast is composed of 7-inch (18-cm) square composite tubing. The SWOT mast is composed of 10-inch (25-cm) squar e composite tubing. Figure 3 displays the spring mechanisms for each mission. The smaller, 7-inch (18-cm) mast cross- section of NISAR became the driving factor in the design of the spring mechanism to maximize mechanical commonality between projects . Common mounting interfaces were design ed for both p rojects . Ultimately, this led to a cylindrical volume allowance of 7 inches (18 cm) in length and 1.75 inches (44 mm) in diameter for the NISAR spring mechanism. Because of the differences between the SWOT and NISAR stowed hinge angles , as well as differences in hinge angles at different locations on each mast , four different torsion spring configurations were developed, each with the spring arms located at different angles relative to each other in the relaxed position. This can also be seen in Figure 3 when comparing both images.

Document

AMS_2020.pdf

229 A) B) Figure 3. View of A) NISAR and B) SWOT spring mechanisms The spring mechanism is required to meet JPL design requirements for mission critical spring design. As such, springs are required to have a minimum no test yield factor of safety ( FS) of 1.50 and an ultimate FS of 1.65. Further more, JPL design principles impose a minimum mechanism torque margin of 100% in worst case environments at end of life. These dri ving requirements meant the torsion springs needed to produce a minimum deployment -direction torque of 28 inch- pounds (3.2 N -m) at hinge closure. A standard round wire 17-7 precipitation -hardened stainless -steel torsion spring would not produce adequate torque in the volume available without violating mission critical factors of safety. Alternative materials such as Elgiloy and MP35N were considered, but all of the vendors considered for fabrication of these springs had a significantly higher volume of experience working with 17- 7 stainless steel and developmental risk was deemed higher with these alternative materials. Therefore, a geometric solution was developed: a rectangular cross section spring to maximize the moment of inertia within the available volume . After developmental fabrication test runs, the spring wire height -to-width ratio selected for the spring cross section was 3.88 :1. This value was determined to be the highest ratio achievable with available CNC spring winding manufacturi ng capabilities . Spring manufacturing still included many challenges given the propensity of the spring wire to rotate about the axis of the wire during winding and inconsistencies in spring back resulting in non- uniform torsion spring inner and outer diameters. The CNC spring winding configuration is shown in Figure 4 . Guide support features were added to the flight spring mandrel design to prevent twist about the axis of the spring wire at either end of each spring. Figure 4. Torsion Spring on CNC Coiling Machine The rectangular cross section caused early manufacturing issues for the flight units . The springs did not initially meet axial length requirements. Further the wire was prone to unexpected twisting during winding. The initial inclination of the team was to attempt to relax the overall spring length requirement, but that would have had significant ripple impacts into the mature design of the hinge and mast structures. To address length requirement non- compliance, t he initially baselined spring with 29 coils was modified to a baseline design of 27 coils. With this change, the spring design violated JPL design requirements minimum factor of safety requirements . Reducing the number of windings increased the stress in the spring. In consultation with JPL materials expert s, material coupon testing for the flight lot of material was conducted

Document

AMS_2020.pdf

230 to establish higher strength allowables for the flight lot of material to address the slight negative strength margins. Ultimately, the final flight springs w ere successfully manufactured with a variation of less than 0.007 inch (178 µm) in diameter and 0.012 inch (305 µm) in length across twenty -eight units . The torsion spring design that was developed met all requirements , as verified via tensile test witness coupons of the material, destructive winding testing, dye penetrant inspection, and other verification techniques. Figure 5. Cross section of NISAR Spring Mechanism Once the spri ng mechanism design solution was reached, prototype units were built . A prototype test program was successfully completed prior to flight hardware fabrication to reduce risk of issues in the flight hinge and latching mechanism development. The prototype test program included both ambient and thermal functional testing and thermal characterization testing on a flight -like hinge fixture. The prototype program did not include vibrational testing or life testing due to programmatic constraints . The lack of these prototype tests prevented design issues described in the next section from being uncovered prior to the qualification t est program. Mechanism Integrati on and Hardware Failure Hardware Failure Upon successful completion of the prototype test program, fabrication of qualification, flight spare, and flight piece parts ensued. Seven SWOT spring mechanisms were assembled with a qualification unit slated for thermal testing to characterize torque output at the worst case cold, ambient, and hot qualification temperatures. The qualification unit was of a SWOT design but was deemed similar to the NISAR design. Therefore, a single qualification unit was used for both missions. Thermal life testing was conducted after the qualification unit had undergone vibrational testing. Thermal test temperatures and vibrational test levels were set to encompass the environments for both missions. During thermal testing, the spring was wound and unwound manually through its operational range of motion using a rotary turn table. T orque output and rotary angle were tracked with a transducer and encoder, respectively. At the qualification hot temperature, hardware failure was observed. From repeated torsion springs c ycling (winding and unwinding), fragmented Teflon Foreign Object Debris (FOD) was generated. This can be seen in Figure 6. The source of the FOD was determined to be from two glass -filled Teflon sleeve bearings in contact with the inner diameter of each spring inside the mechanism. The spring mechanism continued to function and torque performance was not measurably altered by the fragmentation. Upon further investigation it was determined that the sleeve bearing had begun to fail prior to hot thermal testing.

Document

AMS_2020.pdf

231 Figure 6. Image of spring mechanism during thermal testing with evidence of hardware failure circled After the failure, the qualification spring mechanism was disassembled. The r oot cause was identified to be invalid analytical model simplification. The analytical model simplified the torsion spring geometry as a cylinder with uniform inner diameter. The real rectangular cross section torsion springs had slight variations in the i nner diameter between coils, with sharp cutting edges presented to the Teflon bushings during cycling. Therefore, the contact stress in the real hardware at the cutting edge was substantially higher than in the idealized analytical model. In addition, the sleeve bearing had been designed with a helical cut along the axis of the bearing, designed to allow radial compliance as the torsion spring inner diameter changes during winding/unwinding operations . However, the helical cut also drastically reduced axial stiffness of the part. As such, when the spring coils moved axially during winding, the edges of the bushing began to contact each other and plastically deform. These failures can be seen in F igure 7. This failure resulted in the opening of a JPL Problem Failure Report. As such, a technical team was assembled to oversee the investigation and resolution of the failure. Because of the multi -mission applicability of the hardware design, the team was compos ed of representatives from both the SWOT and NISAR projects. Any resulting actions needed to be approved by both missions. There was programmatic motivation to utilize as much of the existing hardware as possible. Figure 7. Images of Teflon sleeve bear ings after thermal testing Mechanisms Modification Ultimately, a solution was developed that replaced the Teflon sleeve bearing with a grease- plated 440C stainless steel sleeve bearing with modified geometry. Over 400 functional cycles and 20 disassembly procedures of the qualification mechanisms were carried out during the hardware failure. Table 1 summarizes the test campaign that was conducted to find a new sleeve bearing design. The fundamental approach was to change one parameter at a time from the original bushing design and evaluate its effect on the health of the component and test performance until an acceptable solution was found. An ac ceptable solution had to simultaneously meet mechanism torque performance needs as well as avoid significant FOD generation or damage to the bushing through three times the planned number of flight unit life cycles .

Document

AMS_2020.pdf

232 Table 1: Summary of hardware failure investigation testing Index Material Bearing Design Test Type Test Result Notes 1 Teflon Helical Cut Vibrational, Thermal, Cycle Life Fail -Initial Failure 2 6061 Torque Characterization Fail -Torque requirement failure 3 304C Torque Characterization Fail - Noise from mechanism -Torque requirement failure - Helical cut deemed unacceptable 4 Bronze Solid Sleeve Bearing Torque Characterization Fail - FOD found 5 440C Torque Characterization Fail - FOD Found 6 Copper Extended, solid sleeve bearing Torque Characterization Fail - FOD found - Torque requirement failure 7 440C Torque Characterization, Cold Pass - Noise witnessed - Good Torque 8 Bronze Torque Characterization Fail - FOD found - Good torque 9 440C Vibe, Cold, Hot, Cycle Life Pass - Full Life Test - Good Torque - Noise witnessed 10 440C Torque Characterization Pass - Confirm lubricant alleviates noise As described previously, torque t esting of the mechanism included using a transducer and encoder to measure torque and rotational position, respectively. This torque testing was carried out for each potential bushing design. If the torque was deemed acceptable, the unit was then disassembled and inspected for any FOD or other potential failures. Figure 8 displays the torque performance of the spring mechanism for the final bushing design in ambient and cold conditions during cycle life testing (defined as at least three times the expected number of mechanism cycles) . Torque performance is seen to degrade up to 1.1 N -m (10 in- lb) over the course of 30 cycles at ambient . Further, torque performance degrades at cold temperature about 1.1 N -m (10 in- lb). Despite performance degradation , torque never violated the 3.2 N -m (28 in-lb) torque requirement . Also notable, the unwinding torque at cold temperature is seen to be nearly constant. This differs from the analytical model of linearly decreasing torque. The cause for near -constant torque is suspected to be internal mechanism friction caused by migration and degradation of lubricant on the bushing as it is cycled . Toward the end of the Problem Failure Report investigation, during final life cycle testing of the hardware, audible sound was observed from the hardware. This sound triggered further investigation and res ulted in the development of an assembly -level relubrication process for the mechanisms. This procedure eliminated the s ource of the concerning noise. The relubrication process seeks to augment lubrication in areas on the sleeve bearing where lubricant may have worn aw ay during mechanical cycling.

Document

AMS_2020.pdf

233 Figure 8. Torque profile of mechanism as it is wound and unwound at ambient and cold temperature. Blue indicates initial cycles, yellow intermediate cycles, and red represents later cycles. Conclusion Following resolution of the hardware failure, all flight spring assemblies have been updated, passed flight acceptance environmental testing, and have been integrated into both SWOT and NISAR flight masts . The mechanisms have successfully been tested at higher levels of assembly and performance is consistent . The spring mechanisms require re lubricat ion every 8 cycles of ground testing based on the process developed in the hardware failure investigation, wh ich is achievable at the integrated level of assembly . Key l essons learned from the development of these torsion springs : • Avoid rectangular cross section springs unless volume limitations necessitate their use. Round wire springs have greater geometric a nd performance consistency and are simpler to analytically model. • Beware of analytical model simplifications that may oversimplify and invalidate the results. • Rectangular cross section springs will twist about the axis of the wire when wound. This twisting needs to be considered when designing any hardware coming in contact with the spring. • Consider both the wound and unwound geometry of the spring during design of the mechanism. • There is high value in a complete mechanical cycle life test at the prototype stage of development . References 1. Paul Lytal and Marcel Renson. “Spacecraft Common Deployable Boom Hinge Deploy and Latching Mechanisms.” Proceedings of the 44th Aerospace Mechanism Symposium, (May 2017), 403- 416. Acknowledgement The research described above was carried out at the Jet Propulsion Laboratory, California Institute of Technology under a contract with the National Aeronautics and Space Administration. © 2020. California Institute of Technology. Government sponsorship acknowledged.

Document

AMS_2020.pdf

235 NSI Performance Improvement through the use of Automation Jason Kozmic*, Bill Gratzl* and Hobin Lee * Abstract Chemring Energetic Devices has transferred the manufacture of the NASA standard initiator ( NSI) from its Torrance, CA facility to its Downers Grove, IL facility and performed re- qualification of the manufacturing process. As part of the transfer effort, the use of automation has been introduced into the process to eliminate operations historically performed by hand which has led to reduced cryogenic function time variability, increased manufacturing throughput, and safer handling of the energetics. Introduction The NASA Standard Initiator (NSI) along with Chemring’s commercial equival ent (PC -23) is a two (2) pin electrically activated, hot -wire, electro- explosive device which provides a source of pyrotechnic energy used to initiate a variety of space mechanisms for use on both satellite and launch vehicle applications. Mechanisms include pyrotechnic valves, separation nuts/ bolts, cable/ bolt cutters, pin pullers and many others. The reliable initiation of these one- time use mechanisms is often mission critical leading , to stringent test requirements being levied upon the manufacture and acceptance of the NSI . Electrically activated initiators are conceptually simple devices but possess manufacturing sensitivities that can have significant effects on the final performance. A typical electric initiator is depicted in Figure 1 and consists of a glass to metal sealed header (with receptacle), a bridgewire welded across the header pins, energetic ignition mix (ZPP typical) consolidated onto the bridgewire, and a welded closure output. When an electrical stimulus is applied to the head er pins (5 a mps, 20 ms typical) the current heats the thin bridgewire which in turn heats the consolidated ignition mix. Once the ignition mix reaches it s auto- ignition temperature the energetics undergo a self -sustaining reaction which produces heat, gas, and hot particles. These thermal outputs are used to ignite secondary energetics in an energetic train/cartridge or can be used perform work in a device without any additional booster . Figure 1. Cross Section of NASA Standard Initiator (NSI) * Chemring Energetic Devices, Downers Grove, IL; jkozmic@ced.us.com Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 20 20 Glass to Metal Seal w/ Receptacle Electrical Bridgewire Energetic Mix (ZPP) Welded Closure

Document

AMS_2020.pdf

236 Chemring Energetic Devices (CED) is currently NASA ’s only qualified manufacturer of the NSI . The all -fire/ no-fire stimulus, function time (application of power to first pressure) , and peak pressure are all critical performance requirements of an electrical initiator and the NSI requires stringent control of these performance characteristics. All can be sensitive to the manufacturing process with the cryogenic (22 K ) firing time , unique to the NSI , being particularly susceptible. Process Automation Historically , the NSI had been produced out of CED’s recently closed Hi -Shear facility located in Torrance, CA. The manufacturing process involved performing critical operations by hand with consistency limited by the manufacturing technician. These critical operations include blending of the energetic material batches, welding the bridgewire, applying wet energetic mix (slurry ) onto the bridgewire , and loading and consolidating powder zirconium potassium perchlorate ( ZPP) into the output charge cup. Performing these critical tasks by hand has limited manufacturing throughput and product consistency, resulting in cryogenic firing performance with variability that put the acceptance of an NSI lot at risk. With the transfer of NSI manufacturing to CED ’s Downers Grove, IL facility , the use of automation has been introduced into the manufacture of the NSI leading to less variability in cryogenic firing performance and increased manufacturing bandwidth. To date, three critical aspects of the manufacturing process have been automated at CED; bridgewire welding, application of slurry to the bridgewire, and mix and blend of the energetics (Z PP). A photograph of CED ’s automated bridgewire welder is provided in Figure 2. Figure 2. Automated Bridgewire Welder The automated bridgewire welder reduces operat or involvement in the process of loading the prepared glass to metal seal headers into the machine and starting the process . The automatic welder spools the <0.005-in (0.13- mm) diameter bridgewire onto the header pins then positions the weld electrodes using a vision system. Once positi oned , the weld force is contro lled with feedback control and once welded, the system severs/discard s the remaining wire pigtails. Using the automated bridgewire welder has shown to consistently produce welds meeting the 0.95 – 1.15 ohms bridg ewire resistance requirements with an average of 1.08 ohms and a standard deviation of 0.018 ohm across four lots as shown in Figure 3. In

Document

AMS_2020.pdf

237 contrast, the previous hand welding operation involved a technician cutting the wire to length and positioning the wire for welding and securing in place by hand using a hold- down fixture. The extra touch time introduces the potential to damage the thin bridgewire ( nicks, dings, etc.) which can affect bridge resistance and create hot spots that can cause no- fire failures when 1A/1W is applied for 5 minutes. Additionally, this process can result in unit to unit variability in the configuration of the welded bridgewire, not present in automated process, which can influence how energetics interface with the bridgewire and ultimately affect cryogenic function times (see Figure 5 ). While the end quality of the hand bridgewire weld ( resistance, all- fire/ no- fire performance) is similar to that of the automated process , the automated process significantly increases the first pass yield reducing the need to rework or scrap in- process hardware. In conjunction with bridgewire welding, the interface between the bridgewire and the energetics as well as the amount of energeti c material applied to the bridgewire have significant effects on the cryogenic function time. Historically the application of wetted energetics (“slurry”) directly to the bridgewire has been performed by hand and been difficult to precisely control. The Do wners Grove manufacturing line has implemented an automated application process which controls the rate and duration of slurry application, tightly controlling the amount and condition of the s lurry applied to the bridgewire. The process has averaged 5.2 m g with 0.6 mg standard deviation across four lots as shown in Figure 3. Figure 3. Automated Process Consistency (Mean w/ Standard Deviation Error Bars ) Bridgewire welding and application of slurry to the bridgewire are critical operations controlling the all -fire/ no-fire current, output pressure performance of the initiator between –162°C and + 149°C as well as the function time performance at cryogenic temperatures as low as 22 K . The automated processes have demonstrated significant improvement in the function time variability at the cryogenic conditions. A comparison illustrating the effects of process automation is presented in Figure 4. A single blended lot of ZPP (lot 13 -44352) was manufactured using four methods ; hand welded with a hand slurry application using the documented Torrance method, hand welded with a hand slurry application using a Downers Grove developed method, hand welded with automated slurry application, and both automated welding and slurry application. These groups were then tested at both 77 K and 22 K. The data demonstrates higher average function times and wider variability when loaded with the heritage hand operations. By implementing either an improved hand or automated slurry application process to the 00.20.40.60.811.2 DJA DJB DJC MTWBRIDGEWIRE RESISTANCE, OHMSBridge Resistance Consistency 00.0010.0020.0030.0040.0050.006 DJA DJB DJC MTWAPPLIED SLURRY, GRAMSSlurry Consistency

Document

AMS_2020.pdf

238 heritage hand welding process, the average function times and the variability were shown to decrease, which indicates the inherent variability of the hand process. These batches still produced flyers which resulted in standard deviations similar to the hand process, which were attributable to inconsistency in the hand welding operation. By adding the aut omated bridgewire weld process to the automated slurry application process, the flyers were eliminated and both average and standard deviation cryogenic function times was significantly improved. Figure 4. Effect of Process Automation on ZPP L/N 13- 44352 The ZPP lot (13- 4432) depicted in Figure 4 was deemed not suitable for loading into production NSIs during batch acceptance testing due to function time variability which is required to have an average of l ess than 1 ms and a 3- sigma limit less than 3 ms. Historically when a blended ZPP lot failed to meet batch acceptance testing, the ZPP lot was scrapped and a new lot blended. The re sults presented in Figure 4, indicate the root cause of the cryogenic funct ion time failure was not the raw material but the process. The test units built with the same ZPP lot, but with the automated processes produced acceptable results, indicating the lot to be suitable for production. Cryogenic function times of NSI test lots spanning 2013 through 2017, manufactured at both CED’s Torrance facility (using hand operations) and CED’s Downers Grove facility (with automated bridgewire welding and slurry application) is presented in Figure 5. Data shows th at lots loaded at CED’ s Downers Grove facility ( DG) had function times ( 0.43 ms and 0. 44 ms mean with 0. 19 and 0. 23 ms standard deviation) which were faster and tighter than the most consistent lot ever fabricated at CED’s Torrance Facility (0.59 ms mean with 0. 25 ms standard deviation manufactured

Document

AMS_2020.pdf

239 in 2013) and had significantly tighter performance when compared to routine production lots loaded in 2016 and 2017. Figure 5. Cryogenic Function Time Comparison In addition to improved cryogenic firing performance, the implementation of automation has allowed CED to increase manufacturing throughput. The automated bridgewire welder allows CED to bridge 100 initiator headers in roughly an hour and a half which previously took two days when this operation was performed by han d. Automated slurry application has increase capacity from approximately 200 pieces per day to 200 pieces in 30 minutes of application time. This increased capability allows CED to reduce lead time of the NSI and other initiators . Automation has also helped CED to manufacture initiators in a safer manner than previously possible at its Torrance facility. The advantage of ZPP as an initiating material, its high sensitivity and rapid self -sustaining reaction, also makes i t dangerous to blend as a raw material. The blending of the ZPP raw materials (fine granular zirconiu m and potassium perchlorate) had historically been done by hand at a remote facility by specific, highly trained technicians. The hand blending operation previously put the technician in close proximity to the volatile raw ingredients and exposes them to a lethal amount of energetics. Chemring is committed to improving safety and has developed automated energetic blending capability , depicted in Figure 6 . It ensures no personnel are present in proximity to the raw materials during blending. Once a blend is completed the equipment di spenses and seals the ZPP into 7- gram velostat pucks for safe handling. ZPP batches blended using the automated energetic blend equipment have passed required differential scanning calorimetry /thermogravimetry analysis (DSC/ TGA) and heat of explosion testing (HOE) required to accept a batch as well as meet cryogenic function time requirements required to be suitable for producti on NSI manufacture. Along wi th automating the blend process, CED is currently working to load NSIs using a robotic manipulator. The robotic manipulator will remotely load ZPP into the NSI charge cup, consolidate it to a feedback - controlled pressure, then confirm and log ZPP charge weights on a piece part basis. Qualification of this 00.511.522.533.544.5 0 20 40 60 80 100 120 140Function Time (ms)-420F Function Time Results for ZPP for test units built at TR and DG TR, Dec 2016" TR May 2017" DG July 2017" TR circa 2013" TR May 2017" DG July 2017"

Document

AMS_2020.pdf

240 effort is set to begin in 2020 and once complete, in conjunction with the automated blend equipment will allow CED to blend and load ZPP without personnel coming into contact w ith the raw materials. Figure 6. Automated ZPP Blend Machine (left) and Robotic Manipulator (right) By implementing key automation processes, CED has improved consistency of the already reliable NSI and tightened performance, increased the manufacturing throughput, and markedly improved the safety by minimizing the exposure of employees to potentially hazardous conditions. Upcoming work with the robotic manipulator will further increase these benefits. More consistent performance, shorter leads times, and reduced scrap will allow Chemring to better service NASA by providing NSIs . In addition to the NSI, CED offers the PC -23 initiator produced on the same manufacturing line as a commercial equivalent. The PC -23 is currently in use on a number of space platforms for many of the prime contractors and those customers are set to realize the benefits implemented on the NSI.

Document

AMS_2020.pdf

241 Microvibrations Modelling and Measurement on Sentinel 4 UVN Calibration Assembly using a Piezoelectric 6 Component Force Dynamometer Benoît Marquet*, Jean- Yves Plesseria *, Jérôme Jacobs * and Christof Sonderegger ** Abstract The Sentinel 4 (S4) Ultraviolet Visible Near -infrared (UVN) instrument is a high- resolution spectrometer embarked on the MTG -S platform. The radiometric accuracy of the UVN instrument relies on regular inorbit re- calibrations using the C alibration Assembly (C AA). This CAA mechanism is comprised of a multi - functional wheel containing calibrated diffusers. Positioning is via a stepper motor and, as the instrument is mounted on the MTG -S platform, i t is imperative that the resultant microvibrations and torques are controlled as the MTG IRS instrument is incredibly sensitive to them. To this end, CSL has developed a Simulink model that inputs a motion profile to simulate the real-world mechanisms, generat ing the torques and microvibrations at the satellite s center of gravity. To validate this prediction, a specially designed Kistler dynamometer allows the actual forces and torques exported by the CAA qualification model to be measured. Measuring with Piezoelectric Dynamometers Microvibrations consist of extremely small accelerations with very low intensities . Measuring them is a very challenging task and methods for properly doing so have only become available in the past few years. Piezoelectric force sensors and dynamometers are ideally suited for this purpose. Their incredibly high span- to-resolution ratio of greater than 100000 is a particular advantage. Thus, it is possible to measure dynamic force changes down to 0.01 N, even when the object being measured weighs more than 10 kg. The static weight can be “eliminated” by resetting the charge amplifier ( this acts like a tar e function, effectively re-zeroing the scale). In addition, the very high stiffness of piezoelectric force sensors permits natural frequencies of 1000 Hz or more. To optimize the system, the measurement setup itself is also highly impor tant. As can be seen in Fig. 1, the dynamometer needs to be mounted on a vibration- isolated table to prevent structure- borne sound and external vibrations affecting the results . These external influences can severely distort measurements as the dynamometer , and the mass of the object being measured, can generate a large acceleration themselves . With a correct set -up, the interference signal can reach levels below 0.01 N or 0.003 Nm (RMS; 3…350 Hz). * Centre Spatial de Liège, Université de Liège, Angleur, Belgium; jjacobs@uliege.be ** Kistler Instrumente AG, Winterthur, Switzerland ; christof.sonderegger@kistler.com Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

242 Figure 1. Microvibrations measurement setup: payload (green), piezoelectric dynamometer (dark blue), stone table (gr ey). An outer dome is used for noise protection. The achievable measurement frequency range is typically from 1 Hz to 350 Hz. The lower limit is defined by the natural frequency of the vibration- insulated stone table, while the upper limit is determined by the natural frequency of the system consisting of the dynamometer itself together with the object being measured. Correct mounting of the measured object to the dynamometer is critical in order to obtain a good measurement result. The dynamometer and the measured object must be fastened with a sufficient number of bolts to ensure a proper mechanical coupling. Last, but not least, s ound and electromagnetic interference (EMI) should be avoided in or near the measurement setup. The dynamometer is connected to the charge amplifier with a special high- insulation cable. Data recording is handled via a laptop and an analog- to-digital converter. New Measuring Trends As discussed previously, t he maximum measurabl e frequency for microvibrations is currently in the range of about 350 Hz. However , higher cut -off frequencies are now increasingly required in order to allow for the measurement of l arger objects. Unfortunately , the standard dynamometer design itself will be a limiting factor in trying to achieve these requirements. Recent standard dynamometers are comprised of four three axis force sensors sandwiched between a top and base plate, each made of steel ( as shown in Fig. 2). Figure 2. Structure of a piezoelectric dynamometer to measure microvibrations with four 3-component sensors

Document

AMS_2020.pdf

243 The dynamometer behaves like a second- order spring/mass system with a dominant n atural frequency and so measurements must be taken well below this natural frequency. Addition of any further mass lowers the natural frequency even more. Therefore, if the dynamometer itself is small, simply mounting a heavy object to investigate it can itself have a major influence and lower the natural frequency of the system (as illustrated in Fig. 3). Figure 3. Natural frequency of dynamometer with and without additional mass So, i f measurements up to 500 Hz are required, the dynamometer should have a natural frequency of >1500 Hz , otherwise resonance will exert too much influence on the measurement signal. Therefore, a high natural frequency is a must . The dynamometer’s size and stiffness ha ve a considerable effect on t he natural frequency. The larger the dynamometer, the heavier the top plate will be – thereby reducing the natural frequency. This effect cannot , unfortunately, be entirely compensated for by increasing the stiffness of the sensors. However, r ecent advanc ements in dynamometer design mean that higher natural frequencies are now possible. Th e consequence of this is that it becomes possible to easily isolate microvibrations and further reduce their causes. In addition to this , new testing requirements are also demanding an increase in the size of the dynamometers in order to test complete s ubsystems and whole small satellites. Comb ined with the higher frequency requirement , this means , with known materials , the design limits have already been reached and any further improvements could only have been incremental while requiring enormous time and cost to achieve. Hence new materials for the dynamometer top plate were considered. Ceramic Top Plate Dynamometer As previously described, the dynamometer size has a critical impact on natural frequency of the unit , and while they help add stiffness ; heavy top plates are especially unfavorable. A s earch for new materials to be considered for the dynamometer top plate showed that ceramic s offer highly advantageous properties (see Table 1).

Document

AMS_2020.pdf

244 Table 1. Steel and ceramic – material properties Steel 17- 4 PH Ceramic (Al2O3) Specific gravity 7.8 3.84 Modulus of elasticity (Young´s modulus) 190,000 MPa 370,000 MPa Tensile strength 1200 MPa 300 MPa Thermal expansion 10.8 µm/m -K 5.7 µm/m -K The l ow specific gravity and a high modulus of elasticity offered by ceramics show a clear benefit , however , the low tensile strength and low thermal expansion would be drawbacks. As seen in Fig. 4, finite element method (FEM) calculations show ed that natural frequencies can be increased by 40% if ceramic top plates with similar dimensions to steel were used. This would lead to significant improvements in microvibration detection. Figure 4. Natural frequency F x,y direction in relation to dynamometer size In such a case, lower strength could be accept able considering the very small forces and loads i nvolved in microvibration measurement s. In order to still allow for a correct mounting of the dynamometer, the steel base plate was retained, as it has no effect on the natural frequency of the dynamometer. Measurements of the natural frequency in the z -direction show ed a very well -defined peak at about 2570 Hz (Fig. 5). In the shear directions (Fx,y), the natural frequency was about 1950 Hz. For comparison, dynamometers with steel top plates reach about 1400 Hz in the shear di rection.

Document

AMS_2020.pdf

245 Figure 5. Natural frequency in F z direction, Ceramic top plate dynamometer type Z21492 However, after further investigation , the low thermal expansion proved to be a problem; despite the FEM calculation, a full validation of this behavior was not possible. So, an extended investigation with experimental specimens was undertaken to ensure that the difference in thermal expansion, between a steel base plate and a n Al2O3 top plate, would not lead to fractures in the latter because of its special structural design. Effects of Dynamometer mounting The dynamometer must be mounted on a stiff bas e, ideally a stone table or steel table (min 10x more mass than dynamometer with test object). Essential is also proper mounting. The dynamometer must be screwed with four screws on the stiff , flat and clean bas e. Incorrect mounting affects natural frequency heavily. The effect of different incorrect mounting was measured at Kistler in our test lab. The dynamometer was mounted on a stiff bas e plate and excited with an impulse hammer in Z –direction.

Document

AMS_2020.pdf

246 Figures 6 and 7. Test set up for natural frequency measurements Table 2. Configurations for natural frequency measurements Figure 8: Configuration 5 is recommended. In the summary below, it clearly shows the highest natural frequency.

Document

AMS_2020.pdf

247 Characterization of the Optical Calibration System for Sentinel 4 UVN Satellite The CSL required a natural frequency greater than 1500 Hz to characterize the optical calibration system of the Sentinel 4 UVN satellite. Kistler’s dynamometer type Z21492 , with a ceramic top plate, was selected due to its large dimensions and high natural resonance frequency. The dynamometer was tested on an insulated stone table and in a sound- insulated area by Kistler . The next step was the validation measurements of subassemblies for the Sentine l 4 satellite. After completion of this validation, CSL will perform further calibration measurements so that they will be able to offer the space community a superior test facility for characterization of microvibration measurements , down to a noise floor of 0.01 mN (Narrow band noise ΔF=1 Hz). S4/UVN Calibration Assembly The Sentinel 4 mission covers the needs for continuous monitoring of Earth atmospheric composition and air pollution using a high- resolution Ultraviolet/Visible/Near -Infrared (UVN) sounder instrument to be deployed on two geostationary MTG -S satellites. The radiometric accuracy of the UVN instrument relies on periodic in- orbit re -calibrations using the UVN Calibration Assembly (UVN CAA). This mechanism has been designed, built and qualified by CSL. The mechanism is composed of a multi -functional wheel with optical diffusers and a mirror that are successively placed in front of the camera during the calibration. The rotation is activated by a stepper motor and controlled by a resolver. Figure 9. S4/UVN CAA MTG -S being an Earth observation satellite with accurate pointing requirements, the exported torques and micro- vibrations that are generated by subsystems can degrade the performance of the MTG instruments. Therefore, the micro- vibrations emitted by the Calibration Assembly during the motion of the mechanism shall be reduced to the minimum achievable. In this goal, the micro- vibrations and exported torques of the Calibration Assembly were es timated using a model and then characteri zed using the Kistler micro -vibrations dynamometer. Micro -Vibrations Modelling To evaluate the micro- vibration levels in the moving mechanism a Simulink model has been created that incorporated the elements that generate the vibrations as well as the main components that influence the

Document

AMS_2020.pdf

248 transmission of vibrations in the mechanism. The elements that generate the micro- vibrations are: • The stepper motor • The bearing • The friction The main elements that influence the transmission of the vibrations are: • The stepper motor controller • The transfer function from the motor to the spacecraft interface; • The lever effect from the spacecraft interface to the spacecraft Cent er of Gravity ( CoG). Figure 10. Micro -vibrations Simulink model The stepper motor controller is a Simulink block that takes in input the motion that the motor should follow as well as the micro- stepping parameters and transforms it into the sine wave that will be the current input of the stepper motor. The stepper motor is another Simulink block that takes in input the current of the two phases and that generates the output torque of th e motor. The modelling also takes into account the detent torque of the stepper motor. The next block of the simulation is the output shaft. This block simulates the behavior of the driving shaft by modelling the resistive torque on the shaft. That includes the friction in the bearing and the rotation inertia of the multi -functional wheel. This block can also take into account some cyclic resistive torque generated by im perfect balls or track. The last block is the exportation of the results in the MATLAB workspace. The results of the Simulink computation are composed of the torques and forces generated by the assembly at the motor location. After the simulation, a Matlab script is r an in the workspace to take into account for the mechanical transfer funct ion from the motor interface to the Calibration Assembly interface. The input transfer functions were previously recovered from the FEM analysis. Once the temporal results are obtained in MATLAB, a frequency analysis is carried out to identify the exported torques and microvibrations in different frequency bands. The final result of the analysis is a temporal response of the injected micro- vibrations at the spacecraft CoG. Micro -Vibrations Optimization Using the produced model, an optimi zation an alysis was carried out to reduce the micro- vibrations exported to the spacecraft. The driving of the stepper motor can be tuned by multiple parameters. The main parameters are the maximum current injected in the motor and the motion profile. These parameters have been analy zed to reduce the micro- vibrations to the minimum achievable.

Document

AMS_2020.pdf

249 The first optimi zation performed was determining the impact of the input current as well as the sensitivity to the detent torque of the stepper motor. It was found that the mi cro-vibrations created by the stepper motor were primarily affected by the ratio between the holding torque (proportional to input current) and the detent torque. The higher the ratio, the lower the micro- vibrations level. The second analysis performed was determining the impact of the driving profile on the generated microvibrations. A micro -stepping strategy was introduced to reduce the exported micro- vibrations but even with a perfect input signal, the micro- vibrations induced by the stepper motor cannot be erased. It was evaluated that having more than 64 µsteps / step does not improve the micro- vibrations behavior. A specific input current profile was also suggested to, in theory; remove the non -uniformity of the output torque and so to provide a constant output torque during the motion. This profile was working well in simulations even in open- loop conditions with external perturbations. But it was evaluated incompatible with the actual electronic control system due to cross use of table lookup with other subsystems. Following the rejection of an advanced current profile, more common profiles were evaluated including a constant velocity profile, a constant acceleration profile and a jerk profile. For a jerk -based profile, a full motion is divided in 7 parts: 4 transitions phases (Tt), 2 constant acceleration phases (Ta) and 1 constant velocity phase (Tv). By tweaking the transition and acceleration phases, the motion profile could be tuned and optimi zed to reduce the exported micro- vibrations during a complete motion. It was found that the best profile to reduce the micro- vibrations is a jerk profile that tends to a constant velocity profile. Tuning the acceleration part of the jerk profile allowed reducing the micro -vibrations during the transient par t of the curve. Figure 11. Jerk and Acceleration definition Micro -Vibrations Setup Once the sensitivity analysis and optimi zation performed on the Simulink model was complete, it was required to compare the results with physical measurements. Therefore, a test campaign has been performed using the Kistler dynamometer developed for this application. Fig. 12 shows the CAA mounted on the micro- vibrations table.

Document

AMS_2020.pdf

250 Figure 12. UVN CAA mounted on µVib table The table is capable of recor ding the forces in three axes as well as the torques around the three axes with a high sensitivity as described in the first part of this article. Being very sensitive, a favorable environment is required to avoid impacting the measurements. The two main noise contributors are the acoustic and the seismic environment. Since a characterization of Flight Models was foreseen, it required working in ISO5 environment which creates additional acoustic noise due to continuous airflow. However, CSL is equipped with vacuum chambers located in ISO5 environment which allows reducing the acoustic noise by closing the chamber during the measurement. Furthermore, each vacuum chamber is equipped with a very stable optical bench that is decoupled from environmental vi brations thanks to a heavy seismic mass. The environment noises shown in Tables 3 and 4 were reached, with the instrument installed on the dynamometer, in Focal 5 and Focal 2 that are two vacuum chambers at CSL. The measurements are from 0 to 500Hz. Table 3. RMS Noise in Focal 2 Axis Noise Focal 2 Fx 2.9E-2 N rms Fy 2.9E-2 N rms Fz 3.7E-2 N rms Mx 3.8E-3 Nm rms My 3.2E-3 Nm rms Mz 1.2E-3 Nm rms Table 4. RMS Noise in Focal 5 Axis Noise Focal 5 Fx 3.5E-2 N rms Fy 4.9E-2 N rms Fz 1.7E-1 N rms Mx 4.4E-3 Nm rms My 2.1E-3 Nm rms Mz 2.2E-3 Nm rms

Document

AMS_2020.pdf

251 Fig. 13 shows an example of the spectral signature of the noise for the Mz axis. 10-2100102 Frequency [Hz]10-810-610-4Amplitude [Nm]Spectrum of Mz Figure 13. Noise spectral signature for Mz It can be seen from the noise level that the goal of 0.01 N is not reached. This is attributed to the noisy environment in the cleanrooms. Furthermore, the local ISO5 airflows were activated during the test in Focal 5 hence the higher noise level seen. On the spectral signature, a peak at 50 Hz is present as well as harmonics. This peak has limited impact on the measurement regarding the UVN mechanism because its amplitude is much lower than the measured microvibrations. Micro -Vibrations Characterization Three models were characterized using the dynamometer. The first model to be tested was the Life Test Model on which an extensive campaign was done to check the impact of the optimization parameters identified during the modelling of the system. Later, two Flight Models were characterized to check that the behavior is repeatable between the different models. The characterization is performed at the interface of the dynamometer by frequency bands for the six degrees of freedom. Once the results are obtained, the response is rotated and translated to the theoretical injection point of the mechanism. From this injection point, the impact of the mechanism on the CoG of the spacecraft is determined. While the forces at the injection point should be used to compute the torques at the CoG, they are discarded in the computation because the levels are within the noise of the dynamometer and when multiplied by the lever arm, it becomes the main contributor to the torques seen at the spacecraft CoG which is unrealistic. In Figures 14, 15 and 16, a comparison between the model and the measurement is shown for the torque around the Z axis of the mechanism. The measurement is shown for three frequency bands from 1.2 to 500 Hz. The measurements appeared to be pretty well correlated with the simulation results. The simulation allows getting more accurate data at frequencies that cannot be measured, e.g. , very low frequency range.

Document

AMS_2020.pdf

252 Figure 14. Simulated (blue) vs measured (green) µVib Figure 15. Simulated (blue) vs measured (green) µVib

Document

AMS_2020.pdf

253 Figure 16. Simulated (blue) vs measured (green) µVib The Life Test Model measurements showed that the Simulink model conclusions in terms of reduction of microvibrations and exported torques were correct except for the impact of the holding torque to detent torque ratio. It was observed that changing the drive current (to vary the holding torque) didn’t visibly change the emitted perturbations of the mechanism. Conclusion The dynamometer developed by Kistler for CSL’s UVN calibration mechanism is of an innovative design that includes a ceramic top table in order to increase the global Eigen frequency of the system and by this way allows measurements into a larger bandwidth. The measurements performed by CSL for the UVN project showed that the actual microvibrations and exported torques have been correctly modelled by the Simulink model and it will allow using the model to extrapolate the predications towards the very low frequencies (< 1 Hz). While the measured noise levels were suff icient for the current project, it is expected to reach a better performance with the micro -vibrations dynamometer. Indeed, the environment of the measurement set -up could not be efficiently optimi zed because of the stringent cleanliness requirements for U VN and therefore additional measurements will be performed inside and outside cleanrooms with improved acoustic environments in order to reduce the noise level to the minimum achievable. Acknowledgement This activity has been performed for a contract bet ween CSL and OHB -M for the design and development of the UVN calibration assembly.

Document

AMS_2020.pdf

255 Lubrication Concept Evaluated for Geared Actuators under Starved Conditions Erik Nyberg*, Jonny Hansen** and Ichiro Minami* Abstract Lubricant starvation leads to the risk of a shift in the lubrication regime from (elasto)hydrodynamic towards boundary conditons. Effective tribofilm formation is essential to limit surface damages in these conditions, but additive technology for space- grade lubricants is lacking. T his work evaluates the feasibility of a novel type of multifunctional ionic liquid lubricant , for use with multiply alkylated cyclopentane (MAC). Actuator gearboxes are operated under starved conditions in nitrogen atmosphere to evaluate the effectiveness of the tribofilm forming lubricant (designated P -SiSO). The effectiveness of P -SiSO was evaluated from macro to micro scale in both surface and sub- surface analysis by use of microscopy (optical , interferometric , SEM) and X -ray microtomography (XMT) , and mechanisms of effective lubrication are discussed. Introduction Conditions faced in robotic space exploration missions pose significant challenges to lubrication of complex mechanisms. Geared actuators operated in low temperatures require extensive preheating before startup [1], but once in operation they may suffer from lubricant starvation due to limited resupply of lubricant to the contact [2]. In vacuum conditions, native oxide layers quickly wear out and if the lubricant does not form a protective tribofilm, there is high risk of seizure. Perf luoropolyethers (PFPE) and multiply alkylated cyclopentanes (MAC) are heritage lubricants used in space applications. They both have benefits and drawbacks; the main benefit being outstanding resistance to outgassing, but their tribofilm forming properties are problematic. PFPE forms iron fluourides in tribocontacts, which prevents seizure but eventually degrades the system autocatalytically [3] . MAC on the other hand is a neat hydrocarbon, and i s not generally tribochemically active. Additives are possible, but finding effective additive s that are miscible and non- volatile is challenging, and few options are currently available. As space exploration missions are demanding increasing performance of mechanisms, new solutions are urgently required. This paper aims to establish the feasibility of using hydrocarbon- mimicking silicate forming ionic liquid (P -SiSO) as triboimproving additive in MAC. Recent work on P -SiSO In our previous work [4], we described the molecular design of a hydrocarbon- mimicking synthetic lubricant composed of a tetraalkylphosphonium cation and trimethylsilalkylsulfonate anion, and found that it provides excellent lubricating performance under boundary lubrication conditions [5] as well as elastohydrodynamic conditions [6]. The hydrocarbon- mimicking structure enables miscibility with a range of hydrocarbon base fluids, while the ionic structure of P -SiSO enables reduced volatility. Surface analysis has shown that the excellent performance correlates with formation of a novel type of tribofilm, mainly based on silicate . Preliminary studies in vacuum tribometers and outgassing tests [7] have produced positive results and therefore the next step is to evaluate the lubricant under increasingly realistic configurations . Materials and Methods In this work, P -SiSO was evaluated in commercial off -the-shelf (COTS) geared actuators. Sintered metal gears, reduced lubricant fill, and reduced temperatures was employed to provoke lubricant starved * Luleå University of Technology, Luleå, Sweden; erik.nyberg@ltu.se ** Scania CV AB, Södertälje, Sweden Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 20 20

Document

AMS_2020.pdf

256 conditions and accelerate damages . The main focus of the experiment is on the lubricant s ability to limit surface and subsurface damage in the gears operated in lubricant starved conditions. Concept lubricant A concept lubricant was prepared by dissolving 0.4 wt% of a hydrocarbon- mimicking ionic liquid (tetraalkylphosphonium trimethylsilaalkylsulfonate) [4] in m ultiply alkylated cyclopentane. Adequate performance with regar ds to thermal vacuum outgassing and solubility was recently demonstrated by the this lubricant [8], which we hereafter designate as P-SiSO . Two reference lubricants were employed during this work ; Syntheti c Oil 2001a , a multiply alkylated cyclopentane supplied by Nye Lubricants, Inc. (Fairhaven, MA), and Fomblin Z25, a perfluoropolyether supplied by Solway S.A. (Brussels, Belgium). The reference lubricants are designated as MAC and PFPE respectively. Neat t etraalkylphosphonium trimethylsilaalkylsulfonate was synthesized by Nisshinbo Holdings Inc. (Tokyo, Japan). Actuator Gear box Lubrication The geared actuators consist of a planetary gearbox (GP32) and a DC -motor (RE30) with encoder and servo controller (ESCON 36/2), all acquired from Maxon Motor AG, ( Sachseln, Switzerland). The gearboxes are 3-stage planetary gearbox es with 51:1 gear ratio, with max continous torque rating of 4.5 Nm . The servo controller can provide a maximum continous current of 2 A at 25 V, which corresponds to a max continous torque of 2.6 Nm, ensuring that the sintered steel gears are not mechanically overloaded. The gearboxes are dissasembled and cleaned, before relubricated with the test lubricants, as shown in Figure 1(a-b). The amount of lubricant applied is significantly reduced in order to provoke starved conditions; the original grease fill of 1.6 g is replaced with 0.060 g of test lubricant (5 µl to each planet gear and 15 µl to output bearing). After applying the test lubricant the gearbox is rotated under zero load at low speed (5 min at 800 rpm followed by 5 min at 4000 rpm) to achieve a consistent lubricant distribution within the gearbox. Geared Actuator Test Rig (GATR) A custom made geared actuator test rig (GATR) , shown in Figure 1(c -d), was designed and manufactured for the purpose of evaluating the lubricants in a component scale experiment. In this setup, the actuator is mounted in a refrigerated chamber filled with N 2 gas. The main purpose is to subject the actuator gearboxes to operat ion in lubricant -starved and oxygen- reduced conditions in order to perform post test damage evaluation and boundary film analysis . The GATR is equipped with a dynamometer and temperature sensors in order to monitor effects on gearbox efficicency and temperature evolution while running the actuator against a braking torque. The efficiency is defined as the ratio of electrical power input, 𝑃𝑃𝑖𝑖𝑖𝑖, to mechanical power output, 𝑃𝑃𝑜𝑜𝑜𝑜𝑜𝑜. The power input can be determined by motor speed, current, and torque constant, while power ouput is determined by output speed and applied brake torque. Figure 1. (a) Dissasembled 3 -stage planetary gearbox. (b) Procedure of appl ying lubricant to planet gear. (c) Atmospheric chamber enclosing actuator and sensors . (d) Overview of Geared Actuator Test Rig.

Document

AMS_2020.pdf

257 GATR test conditions The GATR experiments are started at –20°C in >99% N 2 and run for a duration of 1h at 8050 rpm at a nominal braking torque of 0.8 Nm. Every 5 minutes, the motor is stopped for 10 seconds before ramping the speed back up to 8050 rpm at 1000 rpm/s. This test is repeat ed twice with the second repetition at 1.0 Nm braking torque, giving a total of 1.5 M input pinion revolutions (~30 000 at output). In total, five actuator units are evaluated: 2 units are lubricated with MAC, 2 with P -SiSO, and 1 with PFPE. Units number 2 are used for repetition of the first test with MAC and P -SiSO. Post-test damage analysis After subjecting the gearbox to 1.5 million input cycles , post test damage analysis is performed in three scales; (1) surface macro scale by optical inspection and digital microscop y, (2) surface micro scale by 3 D surface profilometry and scanning electron microscopy with electron dispersive x -ray spectroscopy (SEMEDS), and finally (3) sub- surface microscale analysis by x -ray micro tomography (XMT) . After initial inspection of MAC and P -SiSO lubricated gears, two gear teeth are cut out of a Stage 3 ( S3) planet gear using electric spark erosion to be further analyzed. The thin tribofilms are analyzed by SEM in low voltage high contrast detector mode (vCD) at 3 kV, usi ng a Magellan 400 FEG -SEM (FEI Company, Eindhoven, The Netherlands) . EDS was performed using an X -Max 80 mm2 X-ray detector (Oxford Instruments, Abingdon, UK) operated at 3- 5 kV, which is just enough to detect the elements C, O, Fe, and Si. Finally t he gear teeth are scanned with XMT using a Zeiss Xradia 510 Versa (Carl Zeiss X -ray Microscopy, Pleasanton, CA, USA) , with a resolution of 4 µm per voxel (volume pixel) . Tiff stacking and a Canny method edgedetection algorithm was employed to quantify sub surface damage from XMT data. Results and Discussion Results I – Efficiency and Temperature The actuator efficiency and temperature over a 1 hour test cycle is shown in Figure 1 . Between the three lubricants, a clear trend in both efficiency and temperatures c an be distinguished with efficiency ranking of P-SiSO>MAC>PFPE. As expected, efficiency and temperature are inversely correlated, with high efficiency corresponding to low temperature increase and vice versa. This corresponds to previous model scale tribot ests where variants of P -SiSO has been shown to reduce friction and wear compared to neat PFPEs as well as formulated lubricants [5], [6]. Figure 2. (a) Measured efficiency of actuator setup when lubricated with MAC, P -SiSO, and PFPE respectively. lubricants. (b) Gearbox housing temperature (T 1) and chamber interior temperature (T 2) over 1 h test (500 000 input revolutions). As expected, efficiency is inversely related to increasing gear house temperature (T 2). Results II – Macroscale surface inspection The gearboxes wer e disassembled and inspected after 3h of test . PFPE showed evidence of heavy wear, with large amount of wear particles. Therefore we focused on MAC and P -SiSO. Inspection of MAC Stage 3, Figure 3(a) , revealed dark particles and discolouring of the separator disc (B) . EDS analysis confirmed a layer rich in carbon, indicating indicating decomposition of the hydrocarbon lubricant . Microscopy image of the MAC driven gear show a clear wear pattern, with particle build up towards the root. Inspection of P - SiSO Stage 3, Figure 3(b), did not show any obvious sign of degradation., but microscopy images show a

Document

AMS_2020.pdf

258 blue and purple region above and below the pitch line of the driven gear, indicating tribofilm formation. Regardless of MAC or P -SiSO, the driven gear showed more signs of wear than the driver gear. Figure 3. Inspection of Stage 3 components. (a) MAC show visible lubricant degradation (confirmed by EDS) on stage 3 separator disc. Microscope image indicate wear above and below pitch line. (b) No sign of lubricant degradation in P -SiSO case. Worn area color shift indicates formation of boundary film. Results III – Surface micro scale and elemental analysis Figure 4 displays the surface topography of the driven gears seen in Figure 3 , together with a topography map of an unworn tooth. In the case of MAC and P -SiSO, the measurement was made after cutting the teeth so that the full size of the gear could be scanned. Despite this, very little data is recorded below the pitch line. Above the pitch line, the MAC shows an elliptic region that covers about 1/3 of the gear, whereas the case of P -SiSO is limited to the edges of the gear. The MAC topography has likely also been severly worn below the pitch line, but the large height differential over the gear profile makes it difficult to capture the effect on surface roughness in this area. Figure 4.Gear surfaces as seen by 3d profilometry in 10x objective. Wear patterns of MAC and P -SiSO are compared with New (unworn) surface. Dedendum is mostly out of range because of gear involute profile, but tendency of high wear by MACis seen . Regions I and II are selected for 50x objective evaluation. Increasing the magnification provides insight to the active wear and damage mechanisms. In Figure 5(a) the MAC surface show signs of scuffing, with adhered particles and abrasive marks. In contrast, Figure 5(b)

Document

AMS_2020.pdf

259 show a surface where the original sintered pores remain, but the load bearing patches between pores are very smooth. A surface roughness profile (avoiding pores) reveal that the surface parameters are significantly improved compared to the unworn refer ence, as well as the MAC or PFPE (not shown). Figure 5. Regions I -II of MAC and P -SiSO respectively, with roughness profiles and parameters along x profile. (a) MAC surface show evidence of partial seizures (scuffing) . (b) P -SiSO produce smooth contact patches and retains the porous structure of the unworn sintered material. Region II (P -SiSO) was analy zed by SEM- EDS investigate the lubrication mechanism. A patchy tribofilm is clearly visible when using the low voltage high c urrent detector (vCD). However, the EDS analysis could not confirm the presence of silicate. When comparing to a SRV model tribotest, shown in the top right inset of Figure 6(b), it is clear that the tribofilms have a similar visible appearance. The EDS spectra also show similarities in terms of Fe/O/C proportions, but Si is lacking in the analy zed gear surface. Possibly, the gear tribofilm is too thin to be detectable by EDS, even at the low accelerating voltages of 3- 5 keV used. Figure 6. SEM -EDS analysis of P -SiSO driven gear Region II . (a) Overview indicate gear surface is covered with tribofilm. (b) Gear tribofilm compared to P -SiSO tribofilm generated in ball -on-flat SRV tribotest. Results IV – Sub-surface micro scale analysis The analysis is limited to ~1 mm3 of the gear teeth loc ated at the x -coordinate corresponding to the center of the worn region in Figure 4. The colormap show the frequency of detected edges in the y -z plane. The gray scale refers to the density of the material; bright regions correspond to metal (dense) and dark regions to pores (air). Addendum (i) and dedendum (ii) regions on the driven (N) and driver (R) side of the gear are chosen for comparison. Figure 7(a) show adhered metal, confirm ing scuffing at region (N ii). Interstingly, in the same region there are also sub- surface edges detected, which indicates risk of sub- surface cracks. The

Document

AMS_2020.pdf

260 subset images show possible crack formations at x -coordinate X 1–4. Figure 7(b) shows significantly less activity in the colormap produced by the edge detection algorithm, indicating lower risk of sub- surface cracks. High friction is usually detrimental to sub- surface cracking. In starved lubrication conditions, the efficiency improvement of P -SiSO over MAC is likely attributed to surface friction, and therefore it is reasonable to assume less sub- surface cracking. Figure 7. XMT analysis indicates potential sub- surface cracking. Driven side dedendum (N ii) is critical region. (a) MAC show severe scuffing with large particles adhered at N ii, and evidence of subsurface cracks in the same region (b) P -SiSO show less overall detected edges, and no evidence of scuffing. Conclusions • A hydrocarbon mimicking ionic liquid combined with multiply alkylated cyclopentane (P-SiSO) was evaluated in geared actuators under starved lubrication conditions in N 2 atmosphere. • P-SiSO significantly reduced surf ace and sub- surface damage, while increasing gearbox efficiency. • P-SiSO covered the gear with a thin tribofilm, comparison with model tribotest indicates silicate. • Surface roughness was clearly improved by P -SiSO, which likely improves (micro- )EHL conditi ons. • XMT is well suited for damage analysis of s intered metal gears. T he porous structure is susceptible to sub- surface cracking, which can be distinguished by XMT over the entire gear volume. Acknowledgements The "Austrian COMET -Program" in the frame of K2 XTribology (project no. 849109) and The Taiho Kogyo Research Foundation (TTRF) through the 2019 First Research Grant provided funding of this work. References [1] K. S. Novak, Y. Liu, C.- J. Lee, and S. Hendricks, “Mars Science Laboratory Rover Actuator Thermal Design,” in 40th International Conf erence on Environmental Systems , 2010, pp. 1– 11. [2] D. Suffern and J. Parker, “Developmental Bearing and Bushing Testing for Mars Gearboxes,” Aerosp. Mech. Symp. , vol. 44, pp. 529 –541, 2018. [3] D. J. Carré, “Perfluoropolyalkylether Oil Degradation: Inference of FeF3 Formation on Steel Surfaces under Boundary Conditions,” ASLE Trans. , vol. 29, no. 2, pp. 121– 125, 1986. [4] E. Nyberg, C. Y. Respatiningsih, and I. Minami, “Molecular design of advanc ed lubricant base fluids: hydrocarbon- mimicking ionic liquids,” RSC Adv. , vol. 7, no. 11, pp. 6364– 6373, 2017. [5] E. Nyberg, J. Mouzon, M. Grahn, and I. Minami, “Formation of Boundary Film from Ionic Liquids Enhanced by Additives,” Appl. Sci. , vol. 7, no. 5, p. 433, 2017. [6] J. Hansen, M. Björling, I. Minami, and R. Larsson, “Performance and mechanisms of silicate tribofilm in heavily loaded rolling/sliding non- conformal contacts,” Tribol. Int., vol. 123, pp. 130 –141, 2018. [7] ECSS, “Thermal vacuum outgassing test for the screening of space materials (ECSS -Q-ST-70-02C),” 2008. [8] E. Nyberg, L. Pisarova, N. Dörr, F. Pagano, A. Igartua, and I. Minami, “Silicate- Forming Triboimprovers for Multiply Alkylated Cyclopentane Base Fluids,” in 22nd International C olloquium Tribology , 2020, p. 1.

Document

AMS_2020.pdf

261 Development of a Universal and Scalable Mechanism Control Electronics Configur ed to A pplication Solely by Parameter and Software C onfiguration Udo Rapp* and Juan Camilo Garcia Hernandez * Abstract Modern motor-driven high- precision mechanisms require dedicated control e lectronics in order to achieve their individual function, e.g. , position, velocity or acceleration control. The development of Mechanism Control Electronics (MCE) is therefore often driven by necessary engineering development or optimization tasks for this individual function rather than by economic and life- cycle requirements . The new concept of a universal Mechanism Control Electronics, co- funded by the German Space Agency (DLR), breaks up with the e ngineering optimization approach as generally applied to each specific project and demonstrates that one electronics unit is able to serve numerous mechanism applications at minimized adaptation need. With the presented development of a new control elect ronics , a versatile light-weight, low -power, low - volume and low -cost solution applicable to a large variety of different mechanism control requirements could be realized. Introduction In the past decades, several generations of Mechanism Control Electronics have been developed at Airbus Defence and Space, including Solar Array Drive Electronics , Antenna Pointing Electronics for LEO/GEO as well as Mechanism Control Electronics for various science scanning and pointing applications . Each of these drive electronics has been a highly integrated unit with strong regard to volume/mass and power budget and was adapted to the specific mechanisms and system requirements in order to optimize the equipment function to the dedicated customer needs. The new c oncept breaks with the optimization for each single application and transfers all operation modes, control loops and telecommand/telemetry functions into software instead. In addition, the interfaces to spacecraft and to the mechanism are implemented fully flexible to be universal for different mechanism/actuator types and for different spacecraft bus interfaces. This paper describes the collection of versatility requirements and the development performed on the different Mechanism Drive Electronic m odules with the resulting Electronics Demonstrator Model . The "Application S ummary and Conclusion" section contains the present status of the Universal Mechanism Control Electronics and describes its flexibility with respect to the mechanism m otor characteristic s, command/ telemetry interface and specific function. * Airbus Defence and Space GmbH , Friedrichshafen, Germany udo.rapp@airbus.com, juancamilo.garciahernandez@airbus.com Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

262 Requirements Analysis The development work was started with a requirements analysis to determine the key requirements enabling a Control Electronics to become a universal Mechanism Control Electronics. These requirements include the following major technical key aspects: Primary power bus variability : Capability to operate at all standard primary power bus voltages, i.e., 28V unregulated up to 50V regulated bus . Power Drive capability: DC/DC Converter and motor amplifier shall be scaled to comply with the majority of s pace mechanisms . An open archite cture for equipment with higher power demand shall be granted. Command/Telemetry variability: The unit shall operate at all typical interface topologies, i.e., Mil-Std-1553B, RS -422, SpaceWire, ML16/DS16, etc. Eventual hardware m odifications for in terface- adaptation shall be kept to an absolute minimum. Operational Mode Control: Handling of m echanism operational modes (e.g. , standby, movement, autonomous functions) shall be completely handled in software. Current - and Motion- Contr ol Loops: Transfer of all motor current -, torque- , position- , velocity -control loops into real-time- operating software. Bandwidth target above 100 kHz (bandwidth to be distributed to all control loops) Mechanism Actuator Interface: Interface to brushless -DC Motors, brushed- DC motors, Linear Actuators or Stepper Motor s at only minor and pre- defined hardware adaptation Motor Filtering/Damping: Motor mechanical characteristics differ significantly with respect to motor type (stepper/BLDC/DC) as well as the motor electrical characteristics. It is mandatory that the universal MCE either allows proper EMC filtering / damping for a high range of motors or, as an alternat ive, facilitate adaptation to specific motor characteristics. Position -Sensor Interface: Interf ace provision to analog and digital, incremental and absolute optical encoders. Interface to r eference switches, external and motor internal Hall Sensors. Open Architecture: Clearly defined interfaces to open the universal Mechanism Control Electronics to other or new mechanism and interface types. Requirement Summary : One universal Mechanism Control Electronics shall be capable to serve at least 90% of mechanism targets (major application bandwidth) without or with minimized non- recurring effort in hardware adaptation. The open architecture shall allow the adaptation to the other 10% of upcoming mechanism applications .

Document

AMS_2020.pdf

263 Figure 1: Universal MCE in cold redundant configuration Figure 2: Core and Actuator Frame universal MCE Overall Architecture Housing definition The MCE housing design is based on module frames for each module with separate box outer walls. Flexibility to missions with severe radiation requirements can be achieved by adding additional shielding thickness to these box walls (no change of electronics parts due to changed radiation requirements) . The unit structural analysis shows first resonance frequencies above 490 Hz in all axes with sufficient margin to the expected loads and confirms the selected “module frame approach” . Both main and redundant unit s are integrated within one enclosure, separated by an internal aluminium wall. MCE module distribution Different concepts have been regarded in order to achieve the distribution of functions to the differ ent MCE printed circuit boards. Following the major requirement for useability in a majority of applications, the modules have been separated into • Core Frame with the mechanism controller system and the interface to primary power • Interface Board as a plug- in module to the Core Frame • Actuator Frame with mot or ampl ifiers and position sensor acquisition circuits The electrical interfaces between these modules have been standardized to allow the usage of different Interface Boards or Actuator Frames without design adaptation needed on the Core Frame. Core Frame (DC/DC Converter, Controller System and Software ) DC/DC Converter Typical electronics units on a spacecraft provide dedicated building blocks to realiz e different spacecraft primary power bus voltages . Associated engineering and design effort are frequently required to adapt the unit , e.g., from 28V unregulated towards a 38V regulated power bus.

Document

AMS_2020.pdf

264 In order to avoid this adaptation from program to program, the Core Frame primar y bus interface has to be able to cope with an enveloping input voltage range from 22V up to 5 2V without the need of any commissioning or adjustment need. A new converter architectural concept has been established, which combines the MCE EMC input filter t ogether with sequenti al converters for a ctuator power and internal supply. The galvanic isolation is maintained by this converter concept. Mechanism Controller System Field Programmable Gate Arrays (FPGA) with hard- coded firmware (VHDL) are frequently use d in Mechanism Control Electronics to combine the specific mechanism control and sensor acquisition interfaces with intelligent mechanism mode control and interface handling. In most cases, the motor current controller or sensor analog acquisition circuits are kept analog and are adjusted for different motor and sensor characteristics. For a universal Mechanism Control System, it is mandatory to consequently transfer unit operational functions and parameters into software, together with actuator control loops and motor current controller. The selected architecture achieving a standardize d controller concept is depicted in Figure 3. This complete architecture has been introduced into one FPGA embedded system and completes the Core Frame together with the DC/DC Converter defined above. Figure 3: Mechanism Controller System of Universal M CE Besides the logic blocks for command/telemetry interfaces and the actuators and sensor interfaces , the major constituent is a dedicated processor system consisting of two processors plus a Floating Point Unit (FPU). An ARM -M1 processor is responsible for the mechanism functions and operation, whereas the NanoProcessor together with the FPU performs all control loop tasks in vi rtual real -time. The overall processing performance is 23 MIPs, 10 MFLOPs (32- bit-float) plus 1.11 MFLOPs (trigonometric) which allows a control -loop bandwith of >100 kHz that can be distributed to current, position and velocity control.

Document

AMS_2020.pdf

265 Software The universal MCE s oftware has been developed following a software layer model. This allows introduc tion of new mechanism operational modes as well as additional real -time control loop tasks without need of full software re- development. Basic operational modes are available but additional user - defined modes can be added without restrictions. Figure 4 shows the functional distribution of software modules and control loops to the two processor system s and the interface of new modes/control -loop functions and para meters to the MCE software. The basic software blocks within the ARM M1 Processor and within the nanoProc remain unchanged for all MCE appications. Memory / FPU / Trigonometric / Timers / Interfaces / Actuators / Sensors / ...ARM M1 ProcessornanoProc - Interface TMTC - Mode Control - S/W Up-/Down-load - FDIR - ...All real-time tasks: - Control Loops - Hardware Supervision - Actuator-Control - Sensor read-out - ...Mode ParametersControl ParametersNew ModesNew RealtimeTasks0 RUN1 WAIT2 DECEL LERAT E 3 SEEK 4 STOPSTOP CMD Modeswitch <> 5Start position reached End position reached ?STOP CMDModeswitch <> 5 Modeswitch <> 5 STDBYMODE 1..4 MODE 5STDBY CMD Step 1: Determine Modes of Operation and Control Loop in Simulink Step 2: Transfer into ARM-S/W and realtime Tasks Step 3: Upload to universal MCE, Validation in Equipment Figure 4: Distribution of Software to the two processors

Document

AMS_2020.pdf

266 Interface Module The interface modules consist of simple electrical level shifters to comply with the interface electrical protocol, all logic protocol tasks are performed in the Controller System. A set of different interface modules , e.g., for RS -422 interfaces, Mil-Bus 1553B topologies exist. The interface module has been mechanically sized to allow the application of an additional programmable component (FPGA) to support other interface or bus topologies that have a high performance demand. Flexibility with D edicated Actuator Frame s The biggest challenge in developing a universal Mechanism Control Electronics results from the high variety of available motor types, motor characteristics as well as the sensor type for position, velocity or acceleration sensing. In motor amplifier design, EMC fil tering and motor current damping is a function of motor amplifier topology (linear vers us PWM technology), PWM frequency, max imum motor current amplitude, and required damping over frequency. Besides this, the motor characteristics (inductance, coil resist ance, Back EMF voltage) and proper dimensioning/shielding have significant influence to the EMC behavior of the system and therefore also to the behavior of the required filter . Although one filter design can fit to a range of motor characteristics, adaptation to significantly different characteristics will always be necessary. This is especially the case for low -inductive motor type in highvelocity systems compared to higher inductive motors in low -velocity / high -torque systems. The same is valid for the sensors measuring position and velocity of the mechanical system. These sensors follow the operational and accuracy demands of mechanical systems and typically differ from equipment to equipment.  During the development work it became obvious that one single electrical circuit dedicated and optimized to one specific motor characteristics can never serve completely different motor topology and characteristics.  Sensor acquisition electrical circuits within the universal MCE need the flexitility to adapt to sensor characteristics, sensor type and their required range, resolution and accuracy pending on mechanism topology. In order to maintain the universal MCE approach despite this adaptation need, it has been decided that specific Actuator Boards, carrying the motor amplifier(s) together with the sensor acquisition cirucits, will be mandatory for a universal architecture. Therefore, the interface between Core Frame and Actuator Frame has been designed as an open architecture to ensure compatibili ty to any motor and sensor type. Such different Actuator Frames are now characterized by the motor type / electrical parameters and by the type of sensors used in the mechanism. Once an Actuator Boar d has been developed and qualified for such a combination, r e-use of this Actuator Frame design for other mechanisms with the same or a similar configuration is guarateed due to the control loops being held under software control. Presently, Actuator Frames are in design and commissioning for A. 2-phase Stepper Motor application (range 10- 50 Ω / 10- 100 mH) with Hall Sensor / Switch / Potentiometer sensors B. 3-phase Brushless -DC Motor application (range 5- 15 Ω / 15-50 mH) with digital serial optical encoder Further mechanism characteristics can be adapted by development of the respective Actuator Frames, the open architecture allows quick response time for any type of mechanism / sensor system. Figure 6 shows the concept of different actuator frame s with their standardized electrical and mechanical interface to the Core Module.

Document

AMS_2020.pdf

267 Universal Core -ModuleCollection of Actuator-ModulesSelect Motor-/Sensor-Type Figure 6: Different Actuator Frames can be connected to the universal Core Module Application Summary and Conclusion The concept for the Universal Mechanism Control Electronics and its development has been performed at Airbus Defence and Space GmbH supported by the German Space Agency (DLR). The combination of a Core Frame provid es a universal primary power interface and sufficient processing power for most mechanism applications . Together with standardized control/data interfaces and mechanism characteristics , the goal to serve a high number of different mechanisms with one archite cture has been sucessfully reached. T he universal MCE provides a platform that can be configured in short term in software / parameters and by adaptation of the approriate Actuator Frame. Figure 7 shows the workflow for adaptation of the universial MCE to mechanis m types and mechanism characteristics. Development Status and First Integration Results Two models of the universal MCE have been buil t using the developed technology. One of these models concentrated on validation of electrical design and performance validation with BLDC and DC motors in different control loop scenarios. The second model already implements the mechanical modular architecture in the final MCE housing, the environmental analysis for thermal, structural and radiation have bee n performed on this model. Control Loops for BLDC and DC motors in complex motion profiles have been tested successfully and the f irst demonstration of these models with different mechanism types is planned for the AMS 2020 symposium.

Document

AMS_2020.pdf

268 Universal Mechanic Control Electronics , configured for specific applicationActuator FrameFlexible SC Power -I/F 22V - 52V Data /Control I /F Type „standard -1" Software functions & control loop parametersData /Control I /F Type „standard -2" Data /Control I /F Type „specific“ Stepper Actuator Type „standard 1"Mechanism 1 Sensor „Hall -Effect“ Stepper Actuator Type „standard 2"Mechanism 2 Sensor „Potentiometer“ BLDC Actuator Type „standard 1"Mechanism 3 Sensor „Encoder“ Any Actuator Type „specific 1"Specific Mechanism Sensor „Any“ select select selectstandard plug -in Core Frame Selection of I/F-Boards Selection of Actuator Frames select select select select SADM , 2-phase / 3-phase SADM BLDC (tracking / low noise ) MWI Instrument Scanning Mechanism MWHS Scanning with complex profile Linear Actuator SystemXAA, Data Downlink Pointing Mechanism Stepper Specific Actuator SystemWide Range of Mechanical SystemsSelection of I /F-plug -in and Actuator Frame , Adaptation of Software and Control Loop Parameters Figure 7: Adaptation flow of Universal MCE to m echanism targets

Document

AMS_2020.pdf

269 Eddy Current Effects in Spacecraft Mechanisms Emilia Wegrzyn*, Claudia Allegranza**, Thomas Adam***, Florian Liebold*** and René Seiler*** Abstract The mastering of electromagnetic design features in spacecraft mechanisms , including desired or undesir ed effects due to eddy currents , has become a promising area for a model -based design and development approach. In this frame, multiphysics type investigations have been performed on a simplified experimental test set-up and on flight representativ e reaction wheel assemblies . The test results have been successfully correlate d with simulation output obtained from basic linearized models up to transient nonlinear representations of eddy current effects in complex geometries. The impact of critical parameters like the electrical conductivity of materials as a function of temperature has been particularly studied . Introduction and Motivation Many actuators used in spacecraft applications are based on electromagnetic principles for their function and operation. In this overall context, eddy currents in electromagnetic devices may have a desired effect, for instance when creating a resistive torque in speed regulators (or dampers) as used for controlling the deployment of solar arrays or other spacecraft appendages. I n many other cases (e.g. reaction wheel assemblies involving electric motors) , eddy currents are associated with losses and/or motion resistance , which are normally undesired and to be minimized [1]. However, in comparison to other industrial sectors, space mechanisms are often used for very specialized and one- of-a-kind tasks, relying on very few hardware prototypes and limited testing in the course of their development . Therefore, the understanding & optimization of electromagnetic design features in their interact ion with other physical effects has become a very i mportant objective, which has been the main motivation for the research presented in this paper . Theor etical Framework Eddy currents have been subject of theoretical elaborations and experiments since the 19th century when Michael Faraday and Léon Foucault were working on this topic . They may be regarded as loops of electric current induced within conduct ive materials by a varying external magnetic field. When considering a disk rotating in an air gap between the pole pieces of a magnet, the resulting torque grow s with the angular speed of the dis k, which may be approximated by Eq. 1 within a limited speed range [2]: 𝑇𝑇= 𝜋𝜋 𝑏𝑏 𝑐𝑐2𝑎𝑎2 2∙ �1−𝑟𝑟2 𝑎𝑎2 (𝑟𝑟2−𝑐𝑐2)�∙𝜎𝜎 𝐵𝐵2 𝜔𝜔 (Eq. 1) T … resistive torque due to eddy currents [Nm] ω … angular speed of the disk [rad/s] σ … electrical ( bulk) conductivity of the disk material [S/m] B … magnetic flux density (average) in the air gap [T] a, b, c, r … geometric parameters [m] In fact , the resistive torque has been shown to grow linearly (according to Eq. 1) at low speeds only, levelling before reaching a maximum at medium speeds and finally decreasing at higher speeds [2]. * Surrey Space Centre, Guildford, United Kingdom ** ATG Europe, Noordwijk, The Netherlands *** European Space Agency (ESA/ESTEC), Noordwijk, The Netherlands Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 20 20

Document

AMS_2020.pdf

270 When eddy currents f low in a spinning metallic disk, they generate their own magnetic field that counteracts the external ( source) magnetic field. At sufficiently high speeds, the resulting total magnetic field significantly reduces, and, hence, the eddy current related torque stays well below the linear growth observed at low speeds [3]. Furthermore , a basic model according to Eq. 1 may assume the disk material to feature a constant electrical conductivit y. However, as the temperature rises due to Joule heating, the dis k’s electrical conductivity decreases (in an approximately linear fashion, acknowledging the temperature coefficient of resistance) . For many metals, this coefficient exceeds 0. 004/K , and therefore even small temperature changes considerably reduce the electrical conductivity. Joule heating reduces the eddy currents in the spinning dis k (by Ohm’s law). The consequences are mani fold: Lower eddy currents decrease the resulting Lorentz forces and thus the resistive torque. However, they also generate a lower counter acting magnetic field that causes the total magnetic field to be higher again . Hence, assuming constant speed, the spinning disk system will reach a steady -state equilibrium governed by Maxwell’s equations together with the relevant thermodynamic and mechanical effects and boundary conditions . In the following paragraphs, the analytical and experimental study with focus on the evolution of the resistive torque vs. speed is outlined. Model ling and Simulation of a Simplified Case In the frame of the study, modelling and s imulation has been performed using the software tools ANSYS Maxwell® and C OMSOL Multiphysics®, with controlled modificati on of the parameters under investigation. The model g eometry has matched the relevant parts of the eddy current test bench described in the following paragraph. The size of the air gap has been parameterized for easy adjustment of the geometry. Furthermore , the dis k thickness has been varied. Figure 1 shows the meshed geometry in ANSYS Maxwell® including the materials used. Figure 1 – Model configuration & mesh in ANSYS Maxwell® Many transient (time- domain) simulation runs have been carried out to study the sensitivity to parameter changes ( in particular, dis k thickness, dis k material & air gap dimension ), across the full speed range (0…4000 rpm). The following model parameters have been used for the simulation runs : ● Air gap ( i.e. distance betw een the permanent magnets): 20…80 mm (in 10 mm steps ) ● Materials of the sample disks : alumin um, copper & stainless steel ● Thickness of the sample disks : 2 & 3 mm Permanent Magnets (NdFe35) Vacuum Variable Air Gap Magnet Assembly (iron) Spinning Disk (alumin um, copper, stainless steel)

Document

AMS_2020.pdf

271 Serving as an example, Figure 2 shows the simulation results in terms of resistive torque vs. disk speed. It represents a case with a 2-mm copper d isk in a 40- mm air gap. As already indicated above, a linear trend is observed in the low speed range ( here: below approx. 1000 rpm). At higher speeds and up to 3000 rpm, the torque increases with a lower gradient . After passing a maximum, the torque starts decreasing in the upper speed range. As suggested by theory, at high speeds the induced eddy currents give rise to a secondary magnetic field that opposes the primary one. Whilst at low speeds this effect remains small, the counteracting magnetic field becomes significant and eventually leads to a drop of the resistive torque. In order t o confirm this effect , among others, a dedicated test bench was created. Figure 2 – Resistive torque vs. speed ( simulation results for 2-mm copper dis k, 40-mm air gap) Eddy Current Test Bench The experimental objectives of the Eddy Current Test Bench (ECTB) have been to measure the resistive torque and Joule heating ( due to eddy current dissipation) as a function of the speed of the rotating dis k (driven by a brushless DC motor). Measured values have been correlated with results obtained from two independent models of the test setup ( in Maxwell and COMSOL ), as well as a simplified analytical model. The ECTB has been designed such that the parameters of Equation 1 are easily tweaked (e.g. , air gap size, dis k thickness and dis k material). When the influence of dis k thickness was to be measured, individual disks of identical material and geometry (apart from thickness) have been compared. When the influence of dis k material was to be measured, dis ks of identical geometry have been compared. A simplified geometry has facilitated th e experiments as it has minimized the effect of unknown and unpredictable factors that would otherwise have influenced the test bench results . The ECTB has been designed to provide a constant magnetic field using a variable air gap magnet. The main components of the ECTB (as shown in Figure 3) are: ● Rotating dis k (of different materials and thicknesses) ● Motor drive assembly (brushless DC motor with casing , drive electronics and torque transducer ) ● Variable air gap magnet assembly ● Translation stages (for the variable air gap magnet assembly ), including baseplate -0.18-0.16-0.14-0.12-0.1-0.08-0.06-0.04-0.020 0 1000 2000 3000 4000Torque [Nm] Speed [rpm]

Document

AMS_2020.pdf

272 Figure 3 - ECTB experimental test setup One of the key results of the investigation has been the measurement of the eddy current related torque (see Figure 4) over a speed range sufficiently wide to observe t he predicted torque drop at high speeds. The orange curve represents directly measured resistive torque due to eddy current s (by subtracting the measured non- magnetic los s torque from the total measured torque). The curves obtained through the Maxwell and COMSOL models are shown in green and blue, respectively. As explained below , different conductivity values have also been introduced into the models. Figure 4 - Correlated results of eddy current related torque (2- mm copper dis k, 40-mm air gap) From a multiphysics point of view , the mechanical, electromagnetic as well as thermodynamic effects are strongly intertwined in the case studied. Joule heating (eddy currents dissipating in the dis k) has been shown to directly depend on the torque. Figure 5 (left side) presents data for a configuration of the ECTB for which the air gap has been minimized to 35 mm (reaching motor limits) , in order to maximize the measured torque and Joule heating. 0.000.020.040.060.080.100.120.140.160.18 0 1000 2000 3000 4000 5000Torque [Nm] Speed [rpm]EC measurement Maxwell - 100% sigma Maxwell - 86% sigma COMSOL - 100% sigma COMSOL - 86% sigma Linear modelMaximum Torque Low speed region (appro x. linear) Medium speed region (nonlinear rise to maximum torque) High speed region (torque reduction)

Document

AMS_2020.pdf

273 Figure 5 - Left: average temperature & torque vs. speed (2- mm copper disc, 35- mm air gap) ; Right: measured temperature distribution of spinning dis k (same configuration as left, at 4500 rpm) Temperature measurements have been made using infrared thermometers pointed at a high emissivity face of the spinning dis k (ensured by coating it with matte black paint) , and torque measurements have been conducted in parallel. It is evident in Figure 5 (left side) that the rises in torque and temperature fall together, as predicted. Regarding the evolution of conductivity of the dis k as function of temperature , the largest measured increase above room temperature has been approximately +35 K (= 55°C - 20°C). Figure 5 (right side) shows a thermal image of the ECTB running in this configuration of maximum heating. The annular shape of the temperature distribution corresponds to where most of the heating occurs, namely where the magnetic field has been strongest (directly between the magnet’s pole pieces that can be seen left side of the image). Assuming the temperature coefficient of resistance to be 0. 00404/K for copper , this implies a decrease in conductivity of the spinning copper dis k of more than 14% compared to its value at am bient temperature. The resulting Lorentz force (and so the torque) also drop by a corresponding amount. T hus, thermal effects are not negligible in this context . An adjusted conductivity has been used for the respective simulat ion runs in the Maxwell and COMSOL , shown in Figure 4 as the “86% sigma” curves (the original “100% sigma” curves are for a conductivity that is not adjusted). As can be seen, this has resulted in even closer correlation of the results with measurements. Eddy Current Effects in Reaction Wheels A relevant case where eddy currents are an unwanted side effect can be found in reaction wheels for spacecraft attitude control . They typically consist of an electric motor driving a metallic flywheel with speeds up to 6000 rpm and more. Under the presence of a relatively strong external magnetic field, possibly originating from magnet ic torque rods used as secondary actuators , an additional loss torque due to eddy currents in the rotating flywheel may occur and result in degraded performance of the reaction wheel concerned. In order t o quantify this effect, a dedicated test was devised using the Magnetic Coil Facility at ESA/ESTEC, which features Helmholtz coil s able to generate magnetic fields up to 7.5 mT. A reaction wheel was placed in the cent er of the Helmholtz coils and operated under varying magnetic flux densities. The reaction wheel assembly used comprises a spoked flywheel made from stain less steel. Its maximum motor torque is about 235 mNm, over a speed range up to 2700 rpm. The l oss torque of the reaction wheel assembly is typically in the range of 10 to 1 5 mNm (excluding the effect of any ambient/external magnetic fields ). 0.000.040.080.120.160.200.24 0.010.020.030.040.050.060.0 0 1000 2000 3000 4000 5000 Torque [Nm]Temperature [ °C] Speed [rpm]Average temp. Loss torque Air gap position Spinning disk

Document

AMS_2020.pdf

274 Figure 6 - Left: Test configuration (reaction wheel spin axis orthogonal to the magnetic field vector ) Right: Loss torque vs. ambient magnetic flux densit y For measuring the influence of ambient magnetic fields on wheel performance, the reaction wheel ’s loss torque was measured while generating magnetic fields with varying magnitude (following a specific waveform to ensure that the observed losses are evidently due to the external magnetic field). The test was repeated for magnetic flux densities ranging from 1.0 to 5.0 mT as well as with different reaction wheel orientations, i.e. , spin axis parallel and orthogonal to the magnetic field vector. No measurable effect was observed when the field direction was parallel to the wheel spin axis. When the field was applied in a direction orthogonal to the spin axis, a significant change of th e loss torque was identified , which closely followed the waveform of the external field vs. time . Figure 6 (right side) shows the dependency of the measured loss torque vs. the magnitude of the ambient magnetic field. It can be seen that eddy current related loss torques can rise to levels of >100 mNm, i.e. nearly half the available motor torque. This happened, however , only for a very significant magnetic field of 5 mT , a level which is unlikely to occur in a real flight situation. Moreover, it can be noted that there has been a cubic relationship between loss torque and flux density . This has been a surprising result since, according to Eq. 1, a square relation ship would be expected. It is assumed that the geometry of the flywheel as well as the spatial distribution and direction of the external magnetic field play a critical role in the generated eddy current effects, wh ich is subject to confirmation by ongoing research. Conclusions Eddy current related loss torque effects have been studied in depth, both in terms of simulation and hardware test result s. A nonlinear dependency of the loss torque on relative motion speed has been confirmed, which can be coherently explained by the combined effect of a counteracting magnetic field (generated by the eddy currents ) and a decrease of electrical conductivity due to the rise in temperature by Joule heating. The results obtained and the consistency between simulation output and measurements are promising and give confidence in the fidelity of the analysis tools, particularly when mechanical, thermal and electro magnetic aspects are combined. Ongoing ESA research on the various los s torque components in reaction wheel assemblies and other space mechanisms will definitely benefit from the investigation. It will allow a more accurate predict ion of mechanism performance (especially at higher speeds ), for example in the frame of long- term health monitoring as outlined in [1] . 020406080100120140 1 2 3 4 5Loss torque [mNm] Ambient magnetic field [mT]2700 rpm y = 1.4x^(2.9) 1800 rpm y = =1.2x^(2.8) 1300 rpm y = 0.9x^(3.0) 800 rpm y = 0.8x^(2.9)Reaction Wheel Assembly Helmho ltz Coils

Document

AMS_2020.pdf

275 Acknowledgments The authors gratefully acknowledge the cooperation and support by many technical groups at ESA/ESTEC, in particular by the colleagues of the AOCS & Pointing Systems Section, the EMC & Harness Section, the Materials & Processes Section, the Metrology Laboratory and the Mechanical Workshop. References 1. Bojiloff, D., Häfner, T. & Seiler, R.: “Health Monitoring for Spacecraft Reaction Wheels”, Proc. of the 17th European Space Mechanisms and Tribology Symposium, Hatfield, Sept. 2017. 2. Smythe, W. R.: "On Eddy Currents in a Rotating Disk" , AIEE Electrical Engineering Transactions , Volume: 61, Issue: 9, September 1942 . 3. Gay, Sebastien E. : “Contactless Magnetic Brake for Automotive Applications”, PhD Thesis , May 2005.

Document

AMS_2020.pdf

277 Recovery and Operational Best Practices for Reaction Wheel Bearings Michael J. Dube*, Jeff Fisher**, Stuart Lo ewenthal+ and Peter Ward++ Abstract Left unaddressed, t he observation of sustained high torque signatures in bearing applications ultimately leads to failure. This paper describes v acuum bearing testing of 40 motors fitted with R4 angular contact bearings with 52% and 54% race curvatures , in all steel and hybrid configurations, runni ng in different lubrication regimes. Testing was intended to identify any differences in bearing performance attributed to operation and configuration, and evaluate approaches to recovering bearings in distress with the goal of restoring bearing performance. This paper will present data and describe approaches employed, including resting and heating, that have in some cases resulted in successful bearing recovery, avoiding the onset of hard failure and restoring performance for v arying periods of time. Data addressing p erformance differences between motors equipped with all steel R4 bearings with 52% and 54% curvatures , as well as, performance differences between all steel bearings and their hybrid counterparts will be presented. Introduction Over the past decade, t here have been several reaction wheel assembly (RWA) in -service bearing related failures. The Kepler s pacecraft, launched in March 2009, experienced failures of two of its RWAs in 2012 and 2013 [1]. The Thermosphere Ionosphere Mesosphere Energetics and Dynamics (TIMED) satellite suffered an RWA failure in 2007. The Dawn space probe experienced failures of one of its reaction wheel s in 2010 and another in 2012, and w hen a third reaction wheel stopped working in 2017, Dawn resorted to its hydrazine t hrusters for attitude control [2]. Similarly, the Far Ultraviolet Spectroscopic Explorer ( FUSE ) spacecraft experienced three bearing- related RWA failures [3]. In the commercial sector, a second generation Globalstar satellite failed to spin up one of its RWAs, despite high torque commands . This event was attributed to a stuck bearing, and shortly after regaining control with the three remaining RWAs , a second RWA exhibiting a similar torque signature failed [4]. Efforts to maintain reaction wheel performance and extended life have included attempts at bearing recovery employing techniques such as rest, increasing temperature, and reversing direction. These actions have had limited success with reliance on the remaining reaction wheels as the only recourse. This approach suffers from t he potential for similar life- threat ening bearing related issues. As a resul t, the NASA Engineering and Safety Center initiated a multi -year assessment that included an extensive bearing test program to evaluate the failure modes and effectiveness of potential on- orbit recovery techniques for R 4 bearings experiencing high torque arising from cage instabilities and lubricant depletion leading to failure. This investigation is one of the most comprehensive efforts of its kind with 40 test motors fitted with R4 angular contact duplex bearings in all stainless steel and hybrid (ceramic balls) configurations, tested in hard vacuum at 3 different test speeds and 2 different temperatures. Motors incorporating R4 bearings with 52% and 54% race curvatures were include d because there is anecdotal evidence that tight ball -race curvatures can lead to early failure, a claim that to the best of our understanding has never been documented. In addition, the testing performed in this investigation and its associated findings are of value to designers of bearings for scanners, gimbals, and other rotary spacecraft actuators. * NASA Langley Research Center, Hampton, VA ** Fisher Aerospace, Sunnyvale, CA + Lockheed Martin Space (retired) , Sunnyvale, CA ++ The Aerospace Corporation, El Segundo, CA Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

278 R4 Bearing Life -Testing The initial test plan envisioned testing 40 brushless DC motors in thermal vacuum bell jar chambers that were available at NASA M arshall Space Flight Center (MSFC) [5]. The large number of motors selected was an attempt to obtain statistically relevant life- test data while discerning any effects on bearing life attributed to r ace curvature, speed, temperature, and mode of operation . Additionally, we were interested in evaluating the differences in performance between all steel and hybrid (ceramic balls) bearings. Motor Selection Motor selection was driven by the size of the bell jars and the chiller plates that they could accommodate. Our focus was placed upon procuring vacuum compatible test motors that could easily incorporate our ABEC -7, angular contact , R4 size , 440C bearings and accommodate the existing test rigs at NASA MSFC . Special vacuum and cleanroom proces ses were imposed on the motor structural parts and windings, as well as, the capability to set the same bearing preload regardless of race curvature or bal l material (steel or ceramic). We approached a high- volume production motor supplier willing to accommodate our custom requireme nts and selected 3- phase brushless DC motors with encoders to provide accurate speed control. Shaft and housing were modified for our R4 test bearings to establish proper fits over a test temperature range from ambient to 60 °C. The bearings were preloaded with a wave washer that could be adjusted for each motor to meet the target bearing preload of 2.9 ± 0.23 kg ( 6.5 ± 0.50 lb) using a force gau ge. The 2.9-kg preload approximates the mean contact stress of approximately 8.55e 8 Pa (124 ksi) of flight RWA steel bearings (Figure 1) . The motors had strip heaters with thermocouples mounted on t he rear portion of the housing. Figure 1. Test motor showing component parts (strip h eaters and bearings not shown) . Lubrication There was an expectation that the test motors would have extremely long lifetimes if the bearings were nominally lubricated [6]. In an attempt to minimize test time, the bearings were lubricated with a minimal amount of Nye Synthetic Oil 2001 . The procedure use d to effect this outcome involved immersi ng an oil lubricated bearing in an oil-solvent mixture followed by evaporation of the solvent until a very small meniscus was observed as determined visually under a microscope (Figure 2) . This procedure was referred to as 7% lubricant slosh. The final bearing weights were measured and the variation found to be acceptable and typical of bearing to bearing variation in the field. One exception to performing this procedure was motor 40 which received a 3% lubricant slosh wher e a meniscus was not observed. This exception was to

Document

AMS_2020.pdf

279 reproduce the lubrication condition present in a prototype test that was run previously in an all steel configuration. Figure 2. Oil meniscus present in R4 Bearing after 7% lubricant slosh ( Nye Synthetic Oil 2001/heptane) . Figure 3. Test motors mounted on chiller plate and installed in bell jar. Phase I Testing The motor s were divided into three sets, mounted on chiller plates, and each set placed in a vacuum bell jar (Figure 3). Two bell jars labeled Bell Jars #1 and #2 contained 16 motors each and had all steel bearings , while a third bell jar labeled Bell Jar #3 contained 8 motors and had hybrid bearings. Each bell jar contained motors possessing both 52% and 54% bearing curvatures. Phase I testing consist ed of all motors running

Document

AMS_2020.pdf

280 at 60 °C and 314 rad/s ( 3000 rpm) . These conditions were selected to accelerate the test and introduce a large number of bearing revolutions and associated stress cycles on the lubricant by operating in the boundary lubrication regime. Current from the DC test motors was converted to bearing torque to track and trend the health of the bearings. The baseline torque was comprised of bearing plus motor torque, and was recorded at the beginning of the life test and monitored throughout the testing with any observed increases in torque attributed to bearing drag since the motor torque was constant over time. Phase II Testing Phase II testing was initiated shortly after the failure of motor 22 (52% curvature) which took place at 436 million revolutions (Figure 4) . The temperature control for Bell J ars #1 and #2 was first lowered to a plate temperature of 45°C, temporarily , to avoid any potential damage due to dropp ing the temperature too quickly, then adjusted to 25°C resulting in motor temperatures ranging from 26- 28°C. The final plate temperature of 25 °C was selected to operate the motors at a temperature mor e typical of RWA s in flight . At this point , the motors with all steel bearings were categorized into two groups based on curvature and the operating mode of each group subcategorized and adjusted as follows : 1.) 105 rad/s ( 1000 rpm) no zero crossings, 2.) 105 rad/s ( 1000 rpm) with zero crossings, 3.) 52.4 rad/s ( 500 rpm) no zero crossings, 4.) 52.4 rad/s ( 500) rpm with zero crossings . The motors with hybrid bearings were allowed to continue running at 314 rad/s ( 3000 rpm) and 60° C until they achieved ~5.3 billion revolutions where the temperature and speed on the hybrid motors w as adjusted to 27- 29°C and 105 rad/s ( 1000 rpm), respectively , with pair s running with and without zero crossings . A summary of the testing configuration is shown in Figure 5. T he speed was reduced periodically to 31.4 rad/s ( 300 rpm ) on all motors then ramped up to set speed in order to estimate the coulombic and viscous components of friction comprising the total torque. The viscous component was extrapolated to 314 rad/s (3000 rpm) for the motors running at 105 rad/s (1000) and 52.4 rad/s (500 rpm) . This measurement w as useful in tracking the state of lubrication in the bearings over time since the viscous component of friction would slowly decrease while the coulombic friction would increase, providing an indication that the lubricant was being depleted. Microphones were added to detect any noise associated with cage instabilities. In some cases, as the bearings reached end of life, the total torque would suddenly step- up (Figure 6) with the bell jar microphones recording an i ncrease in the background nois e. One example where the background sound amplitude increased 12 to 20 dB in the 1400 to 2400 Hz range, a signature indicative of a bearing experiencing dry cage instability , is shown in Figure 7. Figure 4. Motor 22 torque signature at point of failure in phase I testing (436 million revolutions). Results The test ing described in this paper has been running for over 4 years. T here have been four 440C and two hybrid bearing failures and one motor with 440 C bearings fully recovered. Hybrid motor 40 testing was suspended when its torque versus speed curve showed no viscous response. All of the steel bearing failures, occurred at less than 1.4 billion revolution s, whereas the hybrid bearing failures in motors 33 and 36 occurred at greater than 5.3 billion revolutions, a significant finding in this work. I t is important to note

Document

AMS_2020.pdf

281 that motor 22, with all steel bearings , and motor s 33 and 36 , with hy brid bearings, were all running under identical conditions (314 rad/s, 60 °C) at the time of their respective failures . See Figure 5 and Table 1. Figure 5. Summary of motors in test and results acquired to date.

Document

AMS_2020.pdf

282 Table 1. Summary of Motor and Bearing Results Steel Bearings Motor Curvature (%) Failure (revolutions) Speed (rad/s) Mode Recovery 22 52 4.36e6 314 (3000 rpm) biased No 8 52 5.79e6 105 (1000 rpm) biased Full 7 52 9.95e6 105 (1000 rpm) biased No 29 52 1.09e9/1.34e9 105 (1000 rpm) zero crossings Yes/No 24 54 1.36e9 52.4 (500 rpm) zero crossings No Hybrid Bearings Motor Curvature (%) Failure (revolutions) Speed (rad/s) Mode Recovery 40 54 3.25e9 314 (3000 rpm) Biased Suspended 36 52 5.30e9 314 (3000 rpm) biased No 33 52 5.37e9 314 (3000 rpm) biased No Hybrid Motors Hybrid motors 33 and 36 were shut down to evaluate recovery techniques . The torque signature and acoustic data for motor 33 were consistent with bearing dr y cage instability as shown in F igures 6 and 7, respectively. One motor was rested for 37 days then ru n another 45 h ours bef ore suff ering a probable cage failure. The other motor with hybrid bearings was also rested for 37 days and then heat soake d at 60 °C for another 30 days when i t failed from probable cage fracture after 7 hours of additional operation (Table 1) . Neither motor with hybrid bearings has been torn down for inspection to validate the failure mode so as not to disturb th e remaining motors under test. Figure 6. Motor 33 torque signature for dry cage instability.

Document

AMS_2020.pdf

283 Figure 7. 12-20 dB spikes between 1400- 2400 Hz associated with dry cage instability on motor 33. As mentioned earlier , and as shown in Table 1, motor 40 with hybrid bearings was suspended at 3.25e9 revolutions. This motor was run at 314 rad/s and 50°C to approximate conditions that were run previously in prototype testing where the motor in an all steel configuration failed at ~250 million revolutions . Note that motor 40 was lubricated with a 3% slosh and did not have a visible meniscus. As can be seen in Figure 8, the Coulombic torque increased as the viscous torques decreased. These coincident events were clear indications that the lubricant was depleted and the bearing was running in a “dry” condition. Further operation likely would have triggered a dry cage instab ility and eventual cage fracture, and it was for these reasons that the test was suspended in order to preserve the motor for eventual teardown and inspection. Figure 8. Plot of viscous and Coulombic torque versus distance.

Document

AMS_2020.pdf

284 Steel Motors The same recovery techniques were applied to motors 7 and 29 with all steel bearing configuration and which failed at approximately 1 billion rev olution s. They were rested 37 days at ambient and then heat soaked. Motor 7 reversed direction 5 times and was rested for a total of 227 days w ith no signs of improvement. Motor 29 was rested for 37 days then heat soaked unpowered at 60 °C for an additional 30 days with the motor recovering for approximately 260, 000 revs before failing again. Remarkably, motor 8 with all ste el bearings showed high torques at 579 M revs and was rested at ambient for 37 days. It then recover ed fully, and is currently healthy at 2.2 B revs while performing zero speed crossings (Figure 9). Figure 9. Typical day before failure (blue), f ailure event (orange) , restart (gold ) and recovery (purple) . Motors in Recovery Protocol Motor 15 exhibited anomalous torque spikes at 2.23 billion revolutions, was rested for 42 days, and recovered for 30 days upon which time torque spikes were observed again and the motor rested and subjected to a heat soak. Motor 16 also exhibited torque spikes at 2.28 billion revolutions at which time it was shut down and rested. Motors 15 and 16 were not considered fail ures as recovery was attempted. In contrast, motor 32 exhibited behavior consistent with dry cage instability and was shut down and subjected to a 37- day rest period. Motors 15, 16, and 32 all possessed 54% curvature. The results of the recovery protocol imposed on motors 15, 16, and 32 were not yet available at the time of writing this paper as shown in Figure 5. Figure 10. Film thickness versus temperature at 314 rad/s and 105 rad/s.

Document

AMS_2020.pdf

285 Conclusions The testing described in this paper will continue since only 6 of th e 40 test motors have failed. P ost-test inspection to confirm f ailure modes will also be performed. At this point in the investigation, some key preliminary findings have emerged (Figure 5 and Table 1): 1) Of the four 440C and two hybrid bearing failures, five of the six failed motors had bearings with 52% race curvature suggesting that tight curvatures, while improving load capacity, may possess shorter life. It has long been suspected that tight curvatures can lead to torque irregularities adversely affect ing bearing lubricant life [7]. While our results are preliminary, t his study may be the first to quantitatively confirm this point. 2) To date, only one out of the four motors that failed with all steel bearings occurred at 52.4 rad/s (500 rpm). The other three failures occurred at 105 rad/s ( 1000 rpm) . Of the 16 motors that have been running at 105 rad/s (1000 rpm), three are currently in recovery protocol, three have failed, and one has fully recovered. There has been only one motor failure out of the 15 motors running at 52.4 rad/s (500 rpm). Although, bearings running at the 105 rad/s ( 1000 rpm) had roughly twice the running distance at the time of failure as their 52.4 rad/s (500 rpm) counterparts, it is interesting to point out that the failures of motors that were running at 105 rad/s (1000 rpm) occur red at total revolutions less than currently observed for the 14 motors still running at 52.4 rad/s ( 500 rpm). B earings running at 52.4 rad/s (500 rpm) have less favorable lubricant film thickness and should theoretically fail sooner. 3) Hybrid bearings exhibited smoother, steadier torque signatures and achieved longer lives compared to the 440C bearings. Of particular note, motor 22 with 52% curvature and all steel configuration failed at 436 million revolutions while running under the same conditions as motors 33 and 36 with hybrid bearings which failed at 5.37 and 5.30 billion revolutions, respectively, providing a direct comparison between the performance of steel bearings and hybrid bearings of equivalent curvature. 4) All hybrid motors achieved an excess of 5 billion revolutions with only two failures, while the remaining 5 hybrid motors are still running. One motor with hybrid bearings, motor 40 (Figure 8), was suspended prior to the onset of failure in order to prevent dry cage instability and eventual cage failure. It is important to note that prior to initiating phase I motor 40 was previously run in prototype testing in an all steel configuration at 105 rad/s (1000 rpm) and 30 °C where it failed ~250 million revolutions and resulted in a broken cage due to dry cage instability. While the conditions of operation differed, the 314 rad/s (3000 rpm) and 50° C conditions selected for motor 40 in phase I and phase II testing approached the same lubrication regime affording an indirect comparison of motors with steel bearings versus hybrid bearings (Figure 10). References 1. Kampmeier, J. L., Larsen, R. J., Miglorini, L.F., Larson, K.A., “Reaction Wheel Performance Characterization Using the Kepler Spacecraft as a Case Study”, 2018 Space Operations Conference, 28 May -1 June, 2018, Marseille, France. 2. Bruno, D., “Contingency Mixed Actuator Controller Implementation for the Dawn Asteroid Rendezvous Spacecraft,” AIAA 201 2-5289, AIAA SPACE 2012 Conference & Exposition, 11- 13 September 2012, Pasadena, CA. 3. Sahnow, D. J. et al, “Operations with the New FUSE Observatory: Three- Axis Control with One Reaction W heel,” SPIE Proceedings, Vol. 6266, July 2006. 4. Hacker, J. M., Goddard, J. L., and Lai, P. C., “Globalstar Second Genera tion Hybrid Attitude Control On Orbit Experience,” AAS 14- 454, 24th AAS/AIAA Space Flight Mechani cs Meeting, Santa Fe, NM , 26-30 January, 2014. 5. McMurtrey, Ernest L. , Lubrication Handbook for the Space Industry. Part A: Solid Lubricants. Part B: Liquid Lubricants. NASA TM- 86556, December 1985. 6. Jones, William R; Jr ., Jansen, Mark J. , ”Lubri cation for Space Applications ,” NASA CR- 2005- 213424, 2005.

Document

AMS_2020.pdf

286 7. Loewenthal, Stuart L., “Two Gimbal Bearing Case Studies ,” Proceedings of the 22nd Aerospace Mechanisms Symposium, May 1988, NASA CP -2506.

Document

AMS_2020.pdf

287 Laboratory Studies of Spacecraft Fluid Lubricant Mobility and Film Thickness Peter Frantz*, James Helt * and Steve Didziulis * Abstract Spacecraft mechanisms often rely upon very small amounts of lubricant to survive in the harsh conditions of space. In critical applications, thin oil films must provide both low drag and longevity, with ball bearings in some devices lasting hundreds of bil lions of loaded mechanical cycles. Our lab has developed several unique test facilities to measure the physical properties of thin liquid lubricant films, and to monitor their performance in realistic bearing level tests, such as ball bearings operating up to 6000 rpm. To understand and optimize lubricant performance, our groups strategy is to correlate bearing level test data with the chemical -physical properties of the lubricant and counter -body interfaces as they evolve during operation . This talk will p rovide an overview of our capabilities and highlight some unique aspects of thin film rheology and additive function. Introduction During the acquisition phase of spacecraft mechanisms, we are often asked to estimate the assets usable lifetime. Similarly, during operation we are asked to assist with extending mechanism life by critically assessing telemetry data that informs to the state of the lubricant . In both cases, mechanism life is usually determined by the life of the lubricant in the critical rolling and sliding interfaces. Once the lubricant is depleted, friction and wear lead to deteriorating performance and failure. This separates space mechanisms from many terrestrial applications, where lubricant may often be reappli ed until failure occurs by other processes, such as rolling contact metal fatigue. For spacecraft mechanisms, retention of lubricant is of paramount importance. When managing these mechanisms during operations, we periodically estimate the relative quanti ty of oil to monitor component health. Bearing drag torque, which is typically proportional to spin motor current, can be separated into three terms: 1) a “Coulombic” term, due to sliding and interfacial slip, 2) a hysteresis term, due to losses in compres sion of the Hertzian contact, and 3) a viscous term, due to displacement of the lubricant in the ball path. This final term, the viscous drag torque, has been shown to scale with viscosity to the 2/3 power [1]. Viscosity decreases with increasing temperature according to a model with functional form similar to the Arrhenius Equation[2 , 3]. Taken together, over small changes in temperature, we find that the dependence of viscous drag on temperature is approximately linear , with the constant of proportionality referred to as the viscous coefficient (Q). This value is used as a relativ e measure of the lubricant in the critical interfaces of a spinning bearing. It is usually large at beginning of life, and tends towards zero late in life. Thus , it can be used as an indicator of bearing health. When monitoring this value in a large population of similar high- speed mechanisms, we have found that (after lubricant run in) the value remains nearly constant for most of life. As the lubricant is depleted, the value gradually falls to zero. If there are two bearings in the component, we find that Q initially falls to half of its saturation value, and then later falls to near zero. A graphic representation of this process is shown in Figure 1 for a typical mechanism. For the first 14 years of operation of this hypothetical mechanism, the viscous coefficient (Q) hovered around 0.2 mA/ °F. During the 7h year, the Q value fell to approximately 0.1 mA/°F, and three years later it fell to near zero. At that point, retainer chatter and instability may begin to oc cur. The late stage behavior is interpreted as sequential starvation of the two bearings. The first one becoming depleted of oil at 7 years, indicated by the viscous drag falling to half of its original value. T he * The Aerospace Corporation, El Segundo, CA Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

288 second bearing was depleted at 1 0 years , where the viscous drag fell to near zero. After the onset of depletion sporadic relubrication events occur due to increased bearing disturbances and temperature. Perhaps the most interesting observation of Figure 1, however, is the relative stability through most of life. Models of lubricant degradation and loss typically imply gradual consumption of oil, yet our observations show a stable quantity at the rolling and sliding interfaces until rapid loss late in life. This trend in Q is interpreted as being due to our imperceptibility of the total amount of lubricant in the bearing, much of which is not actively engaged in the critical interfaces. While this bulk oil is gradually consumed or lost, the oil in the interfaces is maintained at a constant level that is controlled by the combination of mechanical and surface forces acting on the oil. These forces are assumed to be constant during spacecraft operations, so the prevailing thought is that the oil volume does not change and is sustained i n balance with oil reserves resting outside the active tribological interfaces. For example, this may be in the form of grease reserves that are channeled to the sides of the ball path. Figure 1. A graphic representation of the trends in viscous coefficient that are often observed in highspeed spinning devices . A simplified model of the geometry is shown in Figure 2. The Hertzian contact between the ball and the race creates a capillary, which draws oil into a meniscus. At some distance from the contact are reserves of oil (held in grease thickener), and this oil migrates between these two bodies by creeping in a th in film across the bearing surfaces (balls, races, retainer). In addition to the capillary forces of the two bodies, the oil will be subject to mechanical forces (such as centrifugal acceleration) and other surface forces (such as thermocapillarity due to temperature gradients from localized heating). The direction and magnitude of lubricant creep will depend on the balance of these forces. Figure 2. Schematic of an idealized ball/race contact and its relationship with lubricant reserves 0.000.050.100.150.200.25 0 2 4 6 8 10 12 14Absolute Value of Viscous Slope Operational Lifetime (Yrs)Typical Viscous Slope Trend

Document

AMS_2020.pdf

289 This demonstrat es that the sustenance of lubricant in the critical interfaces depends on lubricant mobility on bearing component surfaces. Ultimately, the longevity of critical spacecraft components depends on management of thin films of oil. Understanding the factors controlling these films is essential to prolonging life and responding to flight anomalies caused by lubricant starvation. Thin Film Flow The apparatus shown in F igure 3 A is an adaptation of a device that has been described elsewhere to measure flow of very thin lubricant films [4-10]. In our work, the substrate is a polished steel coupon, with characteristics (composition, roughness, etc.) that are similar to bearing components. A fter a thin oil film (h < 5um) is cast on to a polished steel surface , a jet of nitrogen gas is then directed to the coupon surface causing the oil film to thin over time. F igure 3B shows an optical image of the film during thinning by the N 2 gas jet, where the fluids film thickness ( h) is continuously monitored with an interf erometer down to h ~10nm. The rate at which h declines over time, known as the thinning coefficient ( β, units of nm *s), is proportional to the oil’s viscosity and enables us to determine the viscosity of thin fluid films, a fundamental property that governs the oil’s interfacial -surface mobility. Figure 3. A) Thin Film Flow (TFF) apparatus, B) Thin film interference pattern after blow -off with a jet of nitrogen gas in the TFF . Figure 4 shows an example of the film thickness (h) vs time during flow from the center of the target area. In this case, the sample was a 2- µm-thick film of a multiply -alkylated cyclopentane fluid. We find a rapid decrease in film thickness , with a functional form that appears similar to the film height decreasing in inverse proportionality to the elapsed time of exposure to the nitrogen jet. Figure 5 is the same data, plotted as the inverse film thickness (1/h) vs time . This plot provides a better demonstration of the inverse proportionality, where the slope is th e thinning coefficient, β. We also find that there are two distinct regimes of viscous behavior. The first, with larger film thicknesses and early time of the experiment, represents the bulk properties of the fluid. The second regime represents the interfacial properties of the thin film, where interfacial forces that exist between lubricant molecules and the steel surface influence fluid mobility . In this case, the transition between these two regimes occurs at a film thickness of ca. 45 nm. After the transition to the interfacial regime, the viscosity was observed to increase by approximately 32%.

Document

AMS_2020.pdf

290 Figure 4. Plot of film oil film thickness vs time during blow -off using the TFF Figure 5. Plot of the data shown in Figure 4, using reciprocal of film thickness on the y -axis to show the inverse proportionality with time These changes in viscosity and flow rate are important, because the process of resupply to the tribological contacts may be slowed as lubricant consumption proceeds and the scarcit y of free oil leads to reduced oil film thicknesses. Tribometry and Viscosity of Worn Lubricant Films Another significant reduction in oil mobility during operational use may be caused by changes in oil composition. As the lubricant is worn in a tribological contact due to mechanical stresses and chemical reactions, some molecules are broken into smaller components while others are polym erized into larger

Document

AMS_2020.pdf

291 molecular weight varieties. The lower molecular weight products may evaporate more readily, and the larger products tend to accumulate in the vicinity of the contacts. This occurs with potential increases in the effective viscosity, and may affect both resupply of the contact and elasto- hydrodynamic film thickness. To study the effects of lubricant degradation and changing composition on surface flow in the contact region, we have mechanically worn the lubricant using a ball on disk tribo meter in a vacuum chamber. This inhouse developed instrument, shown in Figure 6, uses steel coupons similar to those shown above [7, 9, 11]. A multiply -alkylated cyclopentane oil with phosphate additives was used at 24° C and background pressure of 5 x 10-6 torr. With a load of 2.6 N , the contact stress was approximately 580 MPa and the diameter of the Hertzian contact area was 0 .6 mm. The rotational speed was 90 rpm (boundary conditions). However, in some cases, a lateral oscillation was applied to lift the contact into EHD conditions. Figure 6. High vacuum ball -on-disk tribometer After running under vacuum for 30 hr , samples were transferred to the TFF apparatus for measurements of vis cosity. Three separate tests are shown in the top of Figure 7. The first two (Un88 and Un89) were performed at low speed, and Un 90 at high speed. In each case the contact path is visible as a distinct line in the oil film, caused by a defect in the otherw ise gradual topography of the film thickness due to pinning and scalloping. After blowing with nitrogen gas for over 1 hour, the displaced oil film patterns are shown in the bottom of Figure 7. In the two low speed examples, we find distinct regions of impeded oil flow that range from 3.1 to 5.2 mm across. These tracks are much larger than the diameter of the Hertzian contact, so they are not caused by changes in the substrate surface roughness. Instead they are related to the size of the ball/flat meniscus, where products of the lubricant degradation have been carried from the Hertzian contact and deposited in the region of oil reflow. In the high- speed case (Un90), bands of impeded oil flow are not easily detected and is likely because the EHD contact was not sufficiently stressful on the lubricant to create substantial wear and lubricant degradation polymerization products.

Document

AMS_2020.pdf

292 Figure 7. TFF measurements of oil mobility after wear in the high vacuum tribometer During the thinning experiments shown in F igure 7, the film thickness was monitored interferometrically and plotted in Figure 8. Here, we compare the unworn base oil (green diamonds) with the low speed (red triangles and blue circles) and high speed ( purple squares ) test results . The results after the high- speed tests are within the experimental uncertainty of the unworn base oil tests. There are no detectable e ffects of running under EHD conditions for the times and pressures that were used. However, the impeded flow after the boundary tests are clearly seen in the lower slope (higher viscosity). Another distinguishing feature is that the slope continued to decrease as the film thickness reduced further in the later portions of the test. This implies that viscosity continued to increase with reduced thickness. Another observation is that the onset of interfacial flow increased from 100 nm with the unworn film to over 180 nm after low -speed testing and is due to the presence of much larger molecular weight species formed during degradationpolymerization. Figure 9 provides a helpful visualization of the distribution of worn lubricant products in the vicinity of the contact. The image on the left shows the contact spot on the ball after testing, and the image on the right shows the sliding contact band on the coupon disk. The Hertzian contact spot, the zone of intimate contact between the ball and flat is at the center of the ring on the left -hand image. The “butterfly wing” lobes are caused by residue in the path of oil flow due to compression of the oil film on the surfaces and distortion of the oil meniscus caused by the dynamics of the sliding contact. The inlet zone of the contact is to the right, and the outlet is on the left. The reflow of oil in this pattern with each ball pass on the disk causes wear products to be swept from the Hertzian contact zone and deposited throughout t he area covered by the base of the meniscus on the flat and the cap of the meniscus on the ball (details below). This results in adlayers of polymerized lubricant in the areas where oil flow is necessary for resupply of oil to the contact. Deposits of polymeric residue may impede flow of oil to the contact, reducing the volume of oil in the inlet zone, resulting in reduced EHD film thickness.

Document

AMS_2020.pdf